SONG POST AND FORAGING LOCATION CHARACTERISTICS
OF BREEDING VARIED THRUSHES IN COASTAL
REDWOOD FORESTS OF NORTHWESTERN CALIFORNIA
by
Maurie Beck
A Thesis
Presented to
The Faculty of Humboldt State University
In Partial Fulfillment
of the Requirements for the Degree
Master of Science
May 1997 SONG POST AND FORAGING LOCATION CHARACTERISTICS
OF BREEDING VARIED THRUSHES IN COASTAL
REDWOOD FORESTS OF NORTHWESTERN CALIFORNIA
by
Maurie Beck
T.Luke George, Chair Date Approved by:
Mark A. Colwell, Committee Member Date
Richard G. Botzler, Committee Member Date
Director, Natural Resources Graduate Program Date
97/W-352/4/30 Natural Resources Graduate Program Number
Approved by the Dean of Graduate Studies
John P. Turner, Jr. Date ABSTRACT
In this study I demonstrated the effectiveness of a hierarchical approach in describing habitat characteristics of song posts and foraging sites used by Varied
Thrushes (Ixoreus naevius). I evaluated characteristics of song posts and foraging locations at four study sites in coastal redwood (Sequoia sempervirens) forests of northwestern California from March - August of 1994 and 1995. I measured a suite of mesohabitat (0.04 ha circular plots) and microhabitat (0.5 m radius) scale attributes centered on occupied and random song post and foraging locations within Varied
Thrush territories.
Varied Thrushes appeared to select song post and foraging locations at multiple scales in a hierarchical process. Ninety-five percent (n = 57) of song posts were in trees or snags. There was no difference in proportion of tree species used as song posts between bird-centered and random locations. The most parsimonious matched-pairs
logistic regression (MPLR) model included distance to water, tree DBH (diameter at
breast height), tree type, slope, and tree density. Male thrushes used song posts with
low foliage density near the top of large DBH conifers, located on steeper slopes,
surrounded by a high density of trees, and centered in drainages closer to water
compared to random locations.
iii Seventy-eight percent (n = 43) of foraging locations were on the ground or on downed logs and 16% (n = 9) were in berry bushes. Varied Thrushes foraged almost exclusively on the ground early in the breeding season. Subsequently, thrushes included fruit (predominantly red huckleberry Vaccinium parvifolium) in their diet after young fledged. Most ground foraging locations were on trails or other disturbed habitat types with little ground foliage density. Though many variables were correlated with ground foraging locations, microhabitat foliage density had the greatest correlation to foraging location and was the only variable included in the final MPLR model. This highlights the importance of measuring habitat characteristics at the correct spatial scale. Also, within foraging locations, thrushes used microsites with much less shrub cover compared to what was available in bird-centered plots. It appears that Varied Thrushes selected foraging sites primarily at the microhabitat, rather than the mesohabitat scale.
Based on my results, a multi-scale, hierarchical methodology should be considered when investigating habitat associations of animals.
iv ACKNOWLEDGMENTS
Grants from The Office of Research and Graduate Studies at Humboldt State
University and Sigma Xi, The Research Society, provided funding for my thesis project.
Additional non-monetary support was contributed by the City of Arcata, Grizzly Creek
State Park, Humboldt State University, Prairie Creek State Park, and Redwood National
Park.
I thank the Department of Wildlife and the Department of Biology at Humboldt
State University for providing financial assistance through teaching and research assistantships. Specifically, I would like to thank Dr. Luke George, Dr. Richard
Golightly, and Dr. Ken Lang.
Special thanks go to Rick Johnson, Dana Johnson, and the crew at Grizzly Creek
State Park for their enthusiastic support of this project. Howard Sakai, the wildlife biologist for Redwood National Park, provided permits and access to the park. Mark
Andre of City of Arcata's Planning Department was very helpful in providing me with the GIS database for the Arcata Community Forest.
Without the help of my field assistants and volunteers I would never have been able to accomplish as much as I did. They were tireless in their efforts. My field assistants included Alden J. Miller, Andrew Pontius, and Michael Beve. My volunteers were Pablo Herrera, Karrie Kritzmacher, Kim Revo, and Karen Danner. Also, special thanks to my friend Kristin Cripe for filling in during crunch time.
My committee members, Dr. Mark Colwell and Dr. Richard Botzler, were always constructive in their suggestions, both during the initial stages of the project and the final editing of the manuscript. Dr. Colwell also insisted on fishing expeditions whenever signs of stress became too pronounced. Dr. John Sawyer, Jr., and Dr. Michael
Messler freely offered their help with vegetation sampling and identification. I thank
Dr. Jerry Allen who introduced me to the "multiple factors and complex interactions" of multivariate statistics. Dr. Todd Arnold and Jeff Dunk made time whenever thorny questions arose. Rob Thompson, the stockroom technician for the Department of
Wildlife and Fisheries, provided my equipment needs.
Dr. T. Luke George is my graduate advisor and my friend. He taught me to think like a wildlife biologist and to ask the good questions. Thank you Luke for always believing in my ability, even when I questioned it.
When I had the crazy idea of going back to school and changing careers, my family and friends were there for me. I thank you all. I especially want to thank my mom for her love and her belief in my crazy idea, and my stepfather, Sam, for his wisdom and humor. From my father came a free ranging mind and the strength and determination to lead my own life. Finally, a special thanks to my godparents, Marion and Maurie Kamins, for the connection I have with the natural world and my first
Peterson Field guide when I was 8 years old.
vi TABLE OF CONTENTS
Page
ABSTRACT iii
ACKNOWLEDGMENTS v
LIST OF TABLES ix
LIST OF FIGURES xi
INTRODUCTION 1
STUDY AREA 4
METHODS 6
Experimental Design 6
Selecting Bird-Centered Locations 6
Selecting Random Locations 7
Microhabitat Variables 8
Mesohabitat Variables 10
Statistical Analysis 11
RESULTS 16
Song Post Locations 16
Foraging Locations 21
vii TABLE OF CONTENTS (CONTINUED)
DISCUSSION 33
Song Posts 33
Foraging Habitat 35
Scale Perspectives 37
LITERATURE CITED 41
PERSONAL COMMUNICATIONS 46
APPENDIXES
A. Common shrubs, ferns, and forbs of redwood forests in Northwestern California 47
viii LIST OF TABLES
Table Page
1 Summary of mean differences (± 2 SE) between paired bird-centered and random sites and results of univariate matched-pairs logistic regression using habitat variables to explain use of song posts by Varied Thrushes in Humboldt County, California, 1994-95. Sample Size = 57 18
2 Results of multivariate matched pairs logistic regression using habitat variables to explain use of song posts by Varied Thrushes in Humboldt County, California, 1994-95. Sample Size = 57 20
3 Summary of mean differences (± 2 SE) between paired bird-centered and random sites and results of univariate matched-pairs logistic regression using microhabitat variables to explain use of foraging locations by Varied Thrushes in Humboldt County, California, 1994-95. Sample Size = 43 24
4 Results of multivariate matched-pairs logistic regression using microhabitat variables to explain use of foraging locations by Varied Thrushes in Humboldt County, California, 1994-95. The analysis yielded two equivalent models. The first table depicts the most parsimonious model, based on the lower AIC (Akaike Information Criteria) value. Sample Size = 43 25
5 Summary of mean differences (± 2 SE) between paired bird-centered and random sites and results of univariate matched-pairs logistic regression using mesohabitat variables to explain use of foraging locations by Varied Thrushes in Humboldt County, California, 1994-95. Sample Size = 43 26
6 Results of multivariate matched-pairs logistic regression using mesohabitat variables to explain use of foraging locations by Varied Thrushes in Humboldt County, California, 1994-95. The analysis yielded two equivalent models. The first table depicts the most parsimonious model, based on the lower AIC (Akaike Information Criteria) value. Sample Size = 43 29
ix LIST OF TABLES (CONTINUED)
7 Results of multivariate matched-pairs logistic regression using variables from previous microhabitat and mesohabitat analyses to explain use of foraging locations by Varied Thrushes in Humboldt County, California, 1994-95. The analysis yielded two equivalent models. The first table depicts the most parsimonious model, based on the lower AIC (Akaike Information Criteria) value. Sample Size = 43 30
8 Standardized differences (± 2 SE) in standard deviations (SD) and mean differences (± 2 SE) in real units between bird-centered foraging location microhabitat (0.5 m circular plots) and mesohabitat (0.04 ha circular plots) attributes used by Varied Thrushes in Humboldt County, California, 1994-95. Sample Size = 43 31
x LIST OF FIGURES
Figure Page
1 Frequency of bird-centered vs. random song post tree species in Humboldt County, California, 1994-95 17
2 Spatial characteristics of song posts in trees used by Varied Thrushes in Humboldt County, California, 1994-1995. A) Vertical position of song posts. A proportion of 1 indicates the top 10 % of trees. B) Height of song posts. C) Foliage density within a 0.5 m radius sphere of song posts 22
xi INTRODUCTION
Song post location for courtship and territorial display, and use of specific substrates for foraging are important aspects of avian ecology (Hunter 1980). Within territories, breeding passerines must fmd suitable areas for both of these functions as well as adequate nest sites if they are to breed successfully. However, the characteristics of each are usually quite different (Hunter 1980, Collins 1981). Song post location is often a function of visibility and conspicuousness (Petit et al. 1988) as well as acoustical transmission (Marten and Marler 1977, Hunter 1980, Shieck 1997). Conversely, foraging microsites frequently change as food resources (Hutto 1981, Brush and Stiles
1986) and predation risk vary (Bouskila 1995, Brown and Morgan 1995). Furthermore, the scale at which a microhabitat is used is not only dependent on the organism itself, but on the type of resource being utilized (Sedgwick and Knopf 1992). Whereas prey items may occur within a very restricted substrate, successful courtship and territorial display by an individual may require a much larger area.
Use of foraging locations is likely the result of innate responses constrained by morphology and behavior (Moermond 1988), and choice based on active exploration of available food resources (Hutto 1985). The degree to which innate responses or active choice are expressed depends on such factors as the scale at which selection is measured
(Petit et al. 1988, Sedgwick and Knopf 1992), whether the bird is a specialist or
1 2 generalist (Recher 1988), time of year, type of habitat available, geographical location
(island vs. mainland), and history (Hutto 1985). Birds are quite plastic in behavior and alter their use of foraging substrate due to changes in location, abundance, type of food items available, and nutritional requirements (Lack 1968, Brush and Stiles 1986, Recher
1988, Greenberg 1990).
The Varied Thrush (Ixoreus naevius) breeds in moist coniferous forests from
Alaska south to northern California (Ehrlich et al. 1988). Territorial males usually sing high in trees and have a brighter plumage than females. This suggests there might be a visual component in the choice of song posts by males or that females have duller plumage due to risk of predation during incubation. Early in the breeding season Varied
Thrushes forage almost exclusively on the ground for invertebrates, which provide the protein needed by the female for egg production and by the nestlings for this period of intense development (Ehrlich et al. 1988). Once young fledge, the diet changes to include more fruit, coinciding with the ripening berry crop (Bent 1949, Ehrlich et al.
1988).
Hurt (1996) found that Varied Thrushes are generally absent from small fragments
(< 16 ha) of mature coastal redwood (Sequoia sempervirens) forest in northern
California. The reasons for this distribution pattern are unclear, but possible explanations include: differences in microhabitat, food availability, predation pressure between large and small fragments, or large area requirements of breeding Varied
Thrushes. Examination of the microhabitat use of Varied Thrushes may provide insight into their breeding distribution. There were two main objectives of this study. First, I 3 examined song post and foraging habitat requirements of Varied Thrushes in northern
California. Second, I analyzed the scale at which Varied Thrushes may be choosing their song post and foraging locations. For this reason I used a hierarchical approach
(Hutto 1985, Petit et al. 1988, Sedgwick and Knopf 1992), incorporating both
mesohabitat (measured within 0.04 ha circular plots) and microhabitat (centered on the song post tree or within 0.5 m radius of the foraging location) variables to characterize song post and foraging locations. The specific objectives of this study were to
determine if there are differences: 1) in habitat characteristics between song post or
foraging locations used by Varied Thrushes and paired random locations, and 2)
between mesohabitat and microhabitat characteristics within occupied song post or
foraging locations. STUDY AREA
I studied Varied Thrushes from March through August of 1994 and 1995 at four
locations in Humboldt County, California. Lost Man Creek (UTM (Universal
Transverse Mercator projection): Zone 10, 414,000 m E, 4,575,500 m N) (5416 ha) and
Prairie Creek State Park (UTM: Zone 10, 414,000 m E, 4,579,500 m N) (6300 ha) are
part of Redwood National Park and are located about 10 km north of Orick. Both are
old-growth coastal redwood (Sequoia sempervirens) forests within 5 km of the Pacific
Ocean. Grizzly Creek Redwoods State Park (UTM: Zone 10, 423,500 m E, 4,481,900
m N) (162 ha) is a drier inland old-growth redwood forest along the Van Duzen River,
2 km west of Bridgeville, and 50 km from the Pacific Ocean. It is on the eastern edge of
the fog belt and represents an ecotone between redwood forests to the west and
Douglas-fir (Pseudotsuga menziesii) forests to the east, sharing many plant species with both habitats. The Arcata Community Forest (UTM: Zone 10, 410,000 m E, 4,525,000 m N) (240 ha) is an 80 year-old second growth coastal redwood forest, 1 km northeast of Arcata and 2 km from Humboldt Bay. It lacks the multi-layered canopy and well developed understory found at the three old-growth locations. The City of Arcata manages the forest for limited timber production, recreational use, and as a semi-natural area for the protection of native flora and fauna.
4 5
The four study sites varied in the amount of relatively flat topography along riverine flood plains and hillsides bisected by steep drainages. Elevation ranged from
10 to 350 m. Cool, dry summers (13 - 18 C) and wet, mild winters (7 - 13 C) characterized the maritime climate of the region. Mean annual rainfall ranges from 96 -
210 cm, with most precipitation occurring from October to April. An extensive fog belt stretches inland 50 - 60 km and moderates temperatures during summer and fall.
Redwood forest canopy (>50%) dominated the four study sites. The major tree
species associates included Douglas-fir, grand fir (Abies grandis), Sitka spruce (Picea sitchensis), western hemlock (Tsuga heterophylla), western red cedar (Thuja plicata),
red alder (Alnus rubra), tan oak (Lithocarpus densiflorus), Pacific madrone (Arbutus
menziesii), bigleaf maple (Acer macrophyllum), and California bay (Umbellularia
californica) (Sawyer and Keeler-Wolf 1995). Appendix A includes a list of the most
common shrubs, ferns, and forbs. METHODS
Experimental Design
I used a paired comparison between bird-centered and corresponding random locations to determine if Varied Thrushes used song posts and foraging locations with attributes that were different from those available at random sites within territories
(Larson and Bock 1986, Holway 1991, Dunk et al. 1997). To resolve the spatial scale at which birds selected foraging and singing locations, I measured habitat in a hierarchical fashion (Hutto 1985, Petit et al. 1988, Sedgwick and Knopf 1992). At a microhabitat scale, I measured variables within 0.5 m radius plots centered on a location. At the mesohabitat scale, I measured variables within a 0.04 ha circular plot (Collins 1981,
Noon 1981, Martin and Roper 1988, Sedgwick and Knopf 1992). In each case I measured habitat within paired random plots at the same spatial scale. I also compared microhabitat to corresponding mesohabitat variables within bird-centered locations to determine the scale at which Varied Thrushes used microsites.
Selecting Bird-Centered Locations
I surveyed for song post and foraging locations during the early morning (0500 -
1100 Pacific Standard Time) and late afternoon (1500 - 2000 PST) from 15 May to 1
August 1994 and 25 March to 1 August 1995. Early in the breeding season, I randomly chose a starting location within one of the four study areas. However, soon after
6 7 surveys began, I was able to recognize individual territories based on locations of singing males, and delineate these on maps of the study areas. Subsequently, I randomly chose territories to begin surveys. Based on Brown's (1985) estimate of home range size (approximately 4 - 20 ha) and personal observations, I considered song post or foraging observations to be independent if they were > 200 m apart. Once I observed a singing male and a foraging location for the male, female, and one juvenile during separate surveys within each territory, no further observations were made to insure there was only one independent observation for each individual. I only used initial detections of bird locations to maintain independence of observations (Bell et al. 1990, Hejl et al.
1990). I found song posts by following singing males until the first visual detection. I found foraging locations by visually searching for foraging birds and flagging their initial foraging attempt. Since thrushes often foraged on trails, I deliberately searched off trails, throughout the forest, so as not to bias my sample. For both song post and foraging locations, I recorded the date, time, location, and sex (adults only) or age
(juvenile vs. adult) of the bird. I measured mesohabitat and microhabitat variables within 2 weeks so that habitat (physical conditions and vegetation phenology) was similar to the time of initial observation.
Selecting Random Locations
For each bird-centered point, I selected paired random locations within territories by generating a random direction and distance (25 - 200 m). For song post locations, if the bird-centered point was in a tree, I centered the random location on a tree nearest the random point. I located the center of the random plot at a random angle around the tree 8 bole and a random distance from the center of the tree to the furthest edge of the foliage.
For foraging locations, if the bird-centered point was on the ground, then the random point was the center of the random plot. However, if the bird-centered point was in a shrub or tree, I centered the random location on a shrub or tree nearest to the random point.
Microhabitat Variables
At each bird-centered and random song post location I recorded the tree species, estimated tree height using a topographic clinometer, and measured tree DBH (diameter at breast height) using a DBH tape. For bird-centered song post locations, I recorded substrate as tree, snag, log, or ground. In intervals of ten percent, I ocularly estimated proportion of foliage density within a 0.5 m radius sphere around the bird (Sakai and
Noon 1991). If the bird was in a tree, I measured bird height and distance from tree top using a topographic clinometer. I then derived bird vertical position by dividing bird height by tree height.
For bird-centered and random foraging locations, I visually estimated, in intervals of ten percent, shrub cover, forb cover, bare ground cover, litter cover, and woody debris cover (classified < 30 cm diameter or > 30 cm diameter) within a 0.5 m radius circular plot frame divided into quarters. I defined shrub cover as vegetation > 0.5 m and < 5 m tall. I defined forb cover as non-woody vegetation < 0.5 m tall. I defined litter cover as non-living vegetation < 5 cm in diameter. I measured litter depth to the nearest 0.5 cm by averaging four random measurements, one from each quarter of the
0.5 m radius circular plot. I measured distance to nearest cover using a tape measure. 9
For birds foraging on the ground, a log, or low shrub, I used a 30 x 50 cm checkered vegetation density board to estimate proportion of foliage density within a 0.5 m radius hemisphere. I placed the board at the center of each point and from a distance of 1 m in the four cardinal directions as well as from above, recorded number of squares occluded by vegetation within the hemisphere, then averaged all five measurements (Martin and
Roper 1988). For foraging locations in tall shrubs or trees, I ocularly estimated foliage density in intervals of ten percent, within a 0.5 m radius hemisphere around the bird.
For ground foraging locations, I measured soil moisture by collecting four random soil samples, one from each quarter of the 0.5 m radius circular plot. I collected samples by first removing forest litter and woody debris from the sampling point. Then I scooped approximately 200 g of soil from the upper 10 cm into an aluminum soil sample can,
immediately secured the lid and taped the seam, and placed the cans in two ziploc
freezer bags. In the lab I weighed each sample to the nearest 0.1 g, placed the sample in
a convection oven at 60° C for 24 hours to insure complete drying (R. T. Golightly,
1994 pers. comm.), then reweighed the sample, and calculated percent soil moisture. In
addition, I recorded the shrub or tree species of all foraging locations in shrubs or trees
and the habitat type in which foraging locations occurred. I categorized habitat types as
undisturbed forest or disturbed habitat, which included trails, roadsides, grassy parks,
and yards. For bird-centered points, I recorded substrate as ground, log, shrub, tree, and
snag. 10
Mesohabitat Variables
Within 0.04 ha circular plots, I recorded number and DBH of each species of tree with a DBH > 10 cm, number of trees of each species with a DBH < 10 cm and > 4 m in height, and number of snags. At the center point of each plot, I measured slope with a topographic clinometer, aspect with a compass, and distance to nearest open water with a 50 m tape measure. Two perpendicular 22.6 m transects, oriented in a random direction, bisected the center of each plot. I used the line intercept method (Bonham
1989) along each transect to measure shrub cover (see definitions in microhabitat variables) and woody debris cover < 30 cm and >30 cm diameter. Beginning at a random point within the first 2 m and proceeding at 2 m intervals on alternating sides of the transect, I measured litter depth to the nearest 0.5 cm, and visually estimated forb cover, litter cover, and bare ground cover (see definitions in microhabitat variables) inside a 20 x 50 cm Daubenmeir plot frame (Bonham 1989). This yielded 22 estimates of each variable, which I then averaged. I used the point intercept method to measure forest vegetation structure by developing foliage profiles within seven vertical layers (0-
1 m, 1-2 m, 2-3 m, 3-5 m, 5-10 m, 10-25 m, and > 25 m) at 33 points within each plot. I collected foliage profile measurements at the center point, 5.65 m from the center point
(16 points), and 11.3 m from the center point (16 points) along sixteen, 11.3 m transects, spaced at 22.5° intervals, radiating from the center point of each plot. I measured the presence of vegetation at 1 m intervals from 0-3 m using a 3 m pole. I used a point intercept periscope densiometer (J. 0. Sawyer, Jr., pers. comm.) to record the presence of vegetation in vertical layers above 3 m. I calculated total foliage density 11 by summing the proportion of live vegetation from each of the seven vertical layers computed for each plot and dividing by seven. I computed foliage height diversity
(HID) from the foliage profile for each plot using the Shannon-Weiner index following
Anderson and Ohmart (1986). I defined canopy cover as the proportion of intercepts within the top two layers (10-25m and > 25m) of the foliage profile. I derived basal area from DBH of all trees > 10 cm DBH (Collins and White 1981). I derived conifer importance by adding the relative density of conifers (number of conifers/number of all trees) and relative basal area of conifers (basal area of conifers/basal area of all trees), then dividing by two (Bonham 1989).
Statistical Analysis
I conducted matched-pairs logistic regression (MPLR) analyses (Hosmer and
Lemeshow 1989) to compare song posts and foraging locations of Varied Thrushes to paired random locations. MPLR offered a statistical methodology consistent with a paired design and allowed me to examine microhabitat selection based on what was available to Varied Thrushes within individual territories. I calculated mean differences between paired bird-centered and random locations for each microhabitat and mesohabitat variable. The mean differences were used in the MPLR analyses. As part of a model building strategy, I first conducted univariate MPLR on each independent variable to determine which variables might be useful in classifying song post or foraging locations occupied by Varied Thrushes (Hosmer and Lemeshow 1989). I included variables that were significant at the 0.25 level in the full model. Hosmer and
Lemeshow (1989) suggested the more stringent significance level of 0.05 might exclude 12 variables that could prove important in a multivariate analysis. This strategy also served as an efficient means of variable reduction. I further reduced the number of variables by generating a correlation matrix and removing highly correlated variables (r > 0.60). I retained the correlated variables that were either the most biologically meaningful or were the most significant in the univariate analyses. I then conducted multivariate
MPLR on all the variables included in the full model. I achieved the most parsimonious reduced model by removing variables in a manual, single-step elimination routine
(Hosmer and Lemeshow 1989) and comparing Akaike Information Criteria (AIC)
(Akaike 1973) values (Lebreton et al. 1992). I also compared the reduced model to one using a stepwise selection procedure. Finally, I assessed model fit by comparing the estimated coefficients and standard errors for each variable in the reduced model with the estimates of those variables in the full model. Large changes in the values are evidence for unstable coefficients, due in part to strong interactions between variables, and a poorly fit model with low explanatory power outside that specific study (Hosmer and Lemeshow 1989).
I used song post tree height, song post tree DBH, tree type (conifer vs. deciduous), mean tree DBH, coefficient of variation of tree DBH, basal area, tree density > 10 cm
DBH, tree density < 10 cm DBH, snag density, conifer importance, canopy cover, FHD, total foliage density, shrub cover, wood cover < 30 cm diameter, wood cover > 30 cm diameter, slope, and distance to water as independent variables to differentiate bird- centered and random song posts. Since 57 of 60 song posts were in trees or snags, I used only those observations in MPLR analyses. 13
Because of the large number of independent variables, I used the basic model
building strategy outlined above to conduct separate analyses on microhabitat and
mesohabitat variables of foraging locations (James and McCulloch 1990). In the first
MPLR, I included the following microhabitat variables: forb cover, bare ground cover,
litter cover, litter depth, soil moisture, shrub cover, wood < 30 cm diameter cover, wood
> 30 cm diameter cover, foliage density, and distance to nearest cover. In the second
MPLR, I included the following mesohabitat variables: forb cover, bare ground cover,
litter cover, litter depth, shrub cover, wood < 30 cm diameter cover, wood > 30 cm
diameter cover, conifer importance, canopy cover, FHD, total foliage density, mean tree
DBH, coefficient of variation tree DBH, basal area, tree density > 10 cm DBH, tree
density < 10 cm DBH, snag density, slope, and distance to water. I then combined the
variables from the reduced microhabitat and mesohabitat models and entered them into a third MPLR to build the final model. I only used foraging locations on the ground or on logs in the analyses (n = 43).
I used standardized differences (means -/- std. dev. for a plot) to evaluate whether the DBH of the song post tree differed from the DBH of trees available within bird- centered plots. I analyzed the mean standardized differences in standard deviation units using a one-sample t-test.
I used the standardized difference analysis described previously to determine if foraging microsites had attributes that were different than those available within each bird-centered plot. I compared whether microsite forb cover, bare ground cover, litter cover, and litter depth were different than a distribution of 22 cover values for each of 14 those variables found within the bird-centered plot. I could not perform standardized difference analysis for shrub and woody debris cover because I estimated microsite cover values within 0.5 m radius plots whereas I used the line intercept method to measure those variables within bird-centered 0.04 ha circular mesoplots. Therefore, the results should be viewed with caution. I used a paired t-test to compare microsite cover values with mean cover values for the plot.
I conducted a contingency table analysis using Cramer's V statistic (Hintze 1996) to determine if there was an association between the species of tree used by Varied
Thrushes as song posts versus randomly chosen trees. Cramer's V statistic ranges between 0 (no relationship) and 1 (perfect relationship), is independent of sample size, and does not require expected cell values of greater than 5. I used chi-square goodness- of-fit tests to determine if vertical position and foliage density of song posts within trees were different than expected. I divided vertical position into five equal intervals of 0-
0.2, >0.2-0.4, >0.4-0.6, >0.6-0.8, >0.8-1.0. I divided foliage density into intervals of 0-
0.1, >0.1-0.25, >0.25-0.5, >0.5-0.75, >0.75-1.0.
I used 2x2 contingency table analysis to determine if thrushes used habitat types, categorized as undisturbed forest or disturbed habitat, in different proportions in bird- centered versus random ground foraging locations. I conducted a contingency table analysis using Cramer's V statistic to determine if Varied Thrushes used specific shrub species as foraging locations in different proportions than shrub species in the random plots. 15
I conducted MPLR analyses using SAS (PROC LOGISTIC; SAS Institute 1995).
Model selection was based on AIC values and not significance of variable parameters
(Lebreton et al. 1992). I performed all other analyses using NCSS 6.0.21 (Hintze 1996).
Results were reported as significant if P < 0.05. RESULTS
Song Post Locations
Fifty-seven of 60 (95 %) song posts were in trees (n = 55, 91.7 %) or snags (n = 2,
3.3 %), two were on logs (3.3 %), and one was on the ground (1.7 %). This provides evidence that Varied Thrushes sang primarily from trees or snags. All singing Varied
Thrushes were males.
I found no association between the species of tree used by Varied Thrushes as song posts versus randomly chosen trees (Cramer's V = 0.44, n = 57) (Figure 1). There was also no significant difference in the proportion of conifers used as song posts compared to random song posts (x2 = 2.67, df = 1, P = 0.10).
Based on the results of univariate MPLR, height and DBH of song post tree, distance to water, slope, tree type (conifer vs. deciduous), tree density, and total foliage density met the criteria (P 0.25) for inclusion in the full model used in the subsequent multivariate analysis of song post versus to random locations (Table 1). However, I removed song post tree height from the final analysis because it was highly correlated
(r = 0.77) with song post tree DBH and was less significant in the univariate analyses
(Table 1).
Song posts tended to be in larger DBH conifers, located on steeper slopes, closer to water, and surrounded by a higher density of trees than random locations in the most parsimonious reduced model (Table 2).
16 Tree species Figure 1. Frequency of bird-centered versus random song post tree species in Humboldt County, California, 1994-95. ABGR = Abies grandis, ACCI = Acer circinatum, ACMA = Acer macrophyllum, ALRU = Alnus rubra, COCO = Corylus cornuta, LIDE = Lithocarpus densiflorus, PISI = Picea sitchensis, POBA = Populus balsamifera, PSME = Psuedotsuga menziesii, RHPU = Rhamnus purshiana, SARA = Sambucus racemosa, SESE = Sequoia sempervirens, TSHE = Tsuga heterophylla, UMCA = Umbellularia californica, SNAG = PSME snag or SESE snag. 17 Table 1. Summary of mean differences (± 2 SE) between paired bird-centered and random sites and results of univariate matched-pairs logistic regression using habitat variables to explain use of song posts by Varied Thrushes in Humboldt County, California, 1994-95. Sample Size = 57.
Variable x Difference ± 2 SE Coefficient ± 2 SE Wald x2 P - value a
Song Post Tree Height (m) 9.00 ± 8.39 0.0186 ± 0.0187 3.95 0.047 0 1
Song Post Tree DBH (cm) 45.07 ± 32.01 0.0072 ± 0.0061 5.60 0.018 * 1
Song Post Tree Type (conifer/deciduous) 0.12 ± 0.07 0.8755 ± 1.0646 2.70 0.10 *
Mean Tree DBH (cm) - 4.75 ± 15.06 - 0.0030 ± 0.0096 0.40 0.53
Coefficient of Variation Tree DBH 0.05 ± 0.11 0.6774 ± 1.3778 0.97 0.32
Basal Area (m2/ha) 4.68 ± 89.45 0.0001 ± 0.0016 0.01 0.92
Density of Trees > 10 cm DBH (/ha) 53.95 ± 71.59 0.0016 ± 0.0022 2.06 0.15 *
Density of Trees < 10 cm DBH (/ha) - 7.02 ± 75.88 - 0.0002 ± 0.0019 0.04 0.85
Snag Density (/ha) 0.88 ± 16.70 0.0004 ± 0.0085 0.01 0.92
Conifer Importance 0.003 ± 0.089 0.0556 ± 1.5990 0.005 0.94
Canopy Cover (%) - 3.33 ± 7.64 - 0.8234 ± 1.9016 0.75 0.39
00 Table 1. Summary of mean differences (± 2 SE) between paired bird-centered/random sites and results of univariate matched- pairs logistic regression using habitat variables to explain use of song posts by Varied Thrushes in Humboldt County, California, 1994-95. Sample Size = 57 (continued).
Variable x Difference ± 2 SE Coefficient ± 2 SE Wald x2 P - value a
Foliage Height Diversity - 0.001 ± 0.010 0.3425 ± 6.8182 0.01 0.92
Total Foliage Density (%) 2.67 ± 1.89 2.6763 ± 3.8796 1.90 0.17 *
Shrub Cover (%) -1.01 ± 8.78 - 0.1884 ± 1.6168 0.05 0.82
Wood < 30 cm Cover (%) 0.80 ± 1.68 4.1660 ± 8.9388 0.87 0.35
Wood > 30 cm Cover (%) 2.34 ± 4.12 2.0566 ± 3.7412 1.21 0.27
Slope (degrees) 3.32 ± 3.16 0.0493 ± 0.0508 3.76 0.052 *
Distance to Water (m) - 29.45 ± 18.71 - 0.0136 ± 0.0103 6.96 0.0084 * a - * indicates variables used in full model. 0 indicates correlated variables (r > 0.60) not included in analysis. Numbers indicate which variables are correlated.
.19 20
Table 2. Results of multivariate matched-pairs logistic regression using habitat variables to explain use of song posts by Varied Thrushes in Humboldt County, California, 1994-95. Sample Size = 57.
2 Variable Coefficient ± 2 SE X P - value
Distance to Water (m) - 0.0198 ± 0.0155 6.46 a 0.011
Slope (degrees) 0.0620 ± 0.0662 3.50 a 0.061
Song Post Tree DBH (cm) 0.0056 ± 0.0070 2.57 a 0.11
Song Post Tree Type (conifer/deciduous) 1.0841 ± 1.3570 2.55 a 0.11
Density of Trees > 10 cm DBH (/ha) 0.0026 ± 0.0034 2.43 a 0.12
b Model c 26.04 < 0.0001 a Wald x2 b -2 Log Likelihood X2 c Akaike Information Criteria (AIC) X24.15 21
Song post tree DBH (microhabitat variable) was a much better explanatory variable of song post locations compared to random locations than mean tree DBH
(mesohabitat variable) (model x2 = 8.081, P 0.0045 and model x2 = 0.406, P< 0.5241, respectively). Varied Thrushes used significantly larger DBH trees ( x ± 2 SE = 46.7 ±
28.2 cm ) for song posts that were almost one standard deviation greater ( x = 0.85 ±
0.36 ) than mean DBH of all trees available within each bird-centered mesoplot
(t = 4.73, P < 0.0001).
Song post locations tended to be higher in trees than would be expected by chance
(x2 = 25.02, d f = 4, P < 0.0001). Twenty-five of 57 (44 %) singing Varied Thrushes were positioned in the highest stratum of the trees (Figure 2A.). Twenty-three of 57
(40 %) song posts were within 20 m of the ground (Figure 2B). A greater proportion of
Varied Thrushes than expected used song posts with low foliage density (x2 = 73.22, df
= 4, P < 0.0001). Twenty-four of 57 (42 %) song post locations were characterized by less than ten percent foliage density within a 0.5 m radius sphere of the bird (Figure 2C).
Foraging Locations
Thirty-one of 55 (56 %) foraging Varied Thrushes were males, 15 (27 %) were females, four (7 %) were juveniles, and five (9 %) I was unable to identify. I also observed eight juveniles or fledglings being fed by adults on five different occasions.
Nineteen of 55 (35 %) foraging locations were in undisturbed forest and 36 (65 %) were in disturbed habitats, within or adjacent to forests, such as trails (n = 19, 35 %), roadsides (n = 12, 22 %), grass park (3 = 5 %), and yards (2 = 4 %). A significantly 22
Figure 2. Spatial characteristics of song posts in trees used by Varied Thrushes in Humboldt County, California, 1994-1995. A) Vertical position of song posts. A proportion of 1 indicates the top 10 % of trees. B) Height of song posts. C) Foliage density within a 0.5 m radius sphere of song posts. 23 greater proportion of Varied Thrushes used disturbed habitat types as ground foraging locations compared to random locations (x2 = 53.55, df = 1, P < 0.0001). Forty of 55
(73 %) foraging locations were on the ground, three (5 %) were on logs, nine (16 %) were in berry bushes, two (4 %) were in trees, and one (2 %) was in a snag. Thus,
Varied Thrushes were primarily ground foragers.
Based on univariate MPLR using only microhabitat variables, bird-centered foraging microsites had less litter cover, litter depth, percent soil moisture, shrub cover, wood < 30 cm diameter cover, foliage density, greater distance to nearest cover, and greater bare ground cover and forb cover than random microsites (Table 3).
In the subsequent multivariate MPLR, I used shrub cover and foliage density in separate analyses because both were positively correlated (r = 0.78). I also removed bare ground cover because it was negatively correlated with litter cover (r = -0.94).
Two equivalent models resulted from the multivariate analysis. Ground foraging microsites 1) were surrounded by 60 % less foliage density within a 0.5 m radius hemisphere, and 2) had 75 % less shrub cover and were 1.3 m further from cover than random microsites (Table 4).
Based on univariate MPLR using only mesohabitat variables, bird-centered plots had less litter cover, litter depth, shrub cover, wood < 30 cm diameter cover, canopy cover, snag density, conifer importance, total foliage density, and greater bare ground cover and forb cover than random plots (Table 5). Like song post locations, bird- centered foraging locations were closer to water than random locations. However, Table 3. Summary of mean differences (± 2 SE) between paired bird-centered and random sites and results of univariate matched-pairs logistic regression using microhabitat variables to explain use of foraging locations by Varied Thrushes in Humboldt County, California, 1994-95. Sample Size = 43.
Variable x Difference ± 2 SE Coefficient ± 2 SE Wald x2 P - value a
Forb Cover (%) 10.32 ± 16.48 0.7175 ± 1.1704 1.51 0.22 *
Bare Ground Cover (%) 23.37 ± 11.74 3.9264 ± 3.0112 6.80 0.0091 0 1
Litter Cover (%) - 24.30 ± 12.30 - 3.3981 ± 2.4870 7.47 0.0063 * 1
Litter Depth (cm) - 2.69 ± 1.42 - 0.2631 ± 0.1828 8.29 0.004 *
Soil Moisture (%) - 6.86 ± 4.06 - 8.0616 ± 5.9774 7.28 0.007 *
Shrub Cover (%) - 75.28 ± 11.04 - 5.2966 ± 3.5828 8.74 0.0031 * 2
Wood < 30 cm Cover (%) - 6.74 ± 4.18 - 11.6855 ± 10.0940 5.36 0.021 *
Wood > 30 cm Cover (%) 0.81 ± 8.96 0.1937 ± 2.1084 0.03 0.85
Foliage Density (%) - 59.67 ± 9.58 - 9.4952 ± 9.0406 4.41 0.036 * 2
Distance to Cover (m) 1.26 ± 0.43 1.7512 ± 1.0150 11.90 0.0006 *
a - * indicates variables used in full model. 0 indicates correlated variables (r > 0.60) not included in analysis. Numbers indicate which variables are correlated. Shrub cover and Foliage Density were included in separate full models. 25
Table 4. Results of multivariate matched-pairs logistic regression using microhabitat variables to explain use of foraging locations by Varied Thrushes in Humboldt County, California, 1994-95. The analysis yielded two equivalent models. The first table depicts the most parsimonious model, based on the lower AIC (Akaike Information Criteria) value. Sample Size = 43.
2 Variable Coefficient ± 2 SE X P - value
Foliage Density (%) - 9.4952 ± 9.0406 4.41 a 0.036
b Model c 49.14 < 0.0001
a Wald X2
b-2 Log Likelihood x2
C AIC = 12.57
2 Variable Coefficient ± 2 SE X P - value
Shrub Cover (%) - 9.5649 ± 1X2986 3.45 a 0.06
Distance to Nearest49.81b (m) - 1.1882 ± 1.8572 1.64 a 0.20
Model 49.81 b < 0.0001
2 a Wald X b -2 Log Likelihood x2 c AIC = 14.1 Table 5. Summary of mean differences (± 2 SE) between paired bird-centered and random sites and results of univariate matched-pairs logistic regression using mesohabitat variables to explain use of foraging locations by Varied Thrushes in Humboldt County, California, 1994-95. Sample Size = 43.
Variable X Difference ± 2 SE Coefficient ± 2 SE Wald x2 P - value a
Forb Cover (%) 7.33 ± 9.56 1.5519 ± 2.1246 2.13 0.14 *
Bare Ground Cover (%) 14.40 ± 7.68 5.9464 ± 4.5948 6.70 0.0096 <> 1
Litter Cover (%) - 15.85 ± 7.52 - 7.0606 ± 5.2234 7.31 0.0069 * 1
Litter Depth (cm) - 2.11 ± 0.96 - 0.4372 ± 0.2820 9.62 0.0019 *
Shrub Cover (%) - 26.66 ± 7.76 - 6.4812 ± 3.5934 13.01 0.0003 *
Wood < 30 cm Cover (%) - 1.61 ± 1.14 - 27.2142 ± 23.3919 5.41 0.02 *
Wood > 30 cm Cover (%) - 0.78 ± 6.24 - 1.4125 ± 5.8412 0.23 0.63 <>
Conifer Importance - 0.09 ± 0.13 - 1.1140 ± 1.6220 1.89 0.17 *
Canopy Cover (%) - 10.51 ± 9.66 - 2.1465 ± 2.1674 3.92 0.048 <> 2
Total Foliage Density (%) - 5.23 ± 2.02 - 4.4654 ± 4.3462 4.22 0.040 * 2, 3
Foliage Height Diversity 0.01 ± 0.02 1.5889 ± 5.8614 0.29 0.59 0 2, 3 Table 5. Summary of mean differences (± 2 SE) between paired bird-centered and random sites and results of univariate matched-pairs logistic regression using mesohabitat variables to explain use of foraging locations by Varied Thrushes in Humboldt County, California, 1994-95. Sample Size = 43 (continued).
Variable x Difference ± 2 SE Coefficient ± 2 SE Wald x2 P - value a
Mean Tree DBH (cm) - 5.50 ± 14.66 - 0.0053 ± 0.0140 0.56 0.45 0 4
Basal Area (m2/ha) - 28.28 ± 77.97 - 0.0009 ± 0.0025 0.52 0.47 0 4
Coefficient of Variation Tree DBH -0.02 ± 0.12 - 0.2343 ± 1.8008 0.07 0.79
Density of Trees > 10 cm DBH (/ha) - 31.40 ± 58.88 - 0.0017 ± 0.0033 1.11 0.29
Density of Trees < 10 cm DBH (/ha) - 22.67 ± 80.64 - 0.0007 ± 0.0024 0.32 0.57
Snag Density (/ha) - 16.86 ± 14.48 - 0.0162 ± 0.0158 4.17 0.041 *
Slope (degrees) -4.03 ± 3.78 - 0.0540 ± 0.0556 3.77 0.052 *
Distance to Water (m) - 32.87 ± 18.09 - 0.0190 ± 0.0135 7.92 0.0049 * a - * indicates variables used in full model. 0 indicates correlated variables (r > 0.60) not included in analysis. Numbers indicate which variables are correlated. 28 unlike song post locations, which were on steeper slopes, foraging locations were on more moderate, gradual slopes compared with random locations (Table 5).
In the subsequent multivariate MPLR, I removed canopy cover and bare ground cover because they were correlated with total foliage density (r = 0.67) and litter cover
(r = -0.96) respectively. Two equivalent models resulted from multivariate analysis. In the first model, ground foraging mesoplots had 27 % less shrub cover and 2.1 cm less
litter depth. The second model included the variables from the first model, and
additionally, foraging mesoplots were 32.9 m closer to water and were on slopes that were 4 degrees less than random mesoplots. (Table 6).
After combining the variables from the previous microhabitat and mesohabitat
models, the final multivariate MPLR analysis of ground foraging locations yielded two
equivalent models. The most parsimonious model was identical to the microhabitat
model containing foliage density. The second model contained mesohabitat litter depth
in addition to the variables from the microhabitat model containing shrub cover and
distance to nearest cover (Table 7). Thus, it appears that fine scale microhabitat
variables have more power in explaining use of foraging locations than larger scale
mesohabitat variables.
Within foraging locations, Varied Thrushes used microsites with significantly less
shrub cover than mean shrub cover for bird-centered plots (Table 8). Though microplot
and mesoplot shrub cover were measured differently, the difference is large enough and
corroborates earlier results. This is evidence that microsite shrub cover provided a
better explanation of foraging locations than shrub cover measured at the mesohabitat 29
Table 6. Results of multivariate matched-pairs logistic regression using mesohabitat variables to explain use of foraging locations by Varied Thrushes in Humboldt County, California, 1994-95. The analysis yielded two equivalent models. The first table depicts the most parsimonious model, based on the lower AIC (Akaike Information Criteria) value. Sample Size = 43.
2 Variable Coefficient ± 2 SE X P - value
Shrub Cover (%) - 5.6938 ± 3.7614 9.17 a 0.0025
Litter Depth (cm) - 0.2465 ± 0.3144 2.46 a 0.12
Model c 30.65 b < 0.0001 a Wald X2 b -2 Log Likelihood x2 c AIC = 33.26
2 Variable Coefficient ± 2 SE X P - value
Shrub Cover (%) - 5.3047 ± 4.4758 5.62 a 0.017
Litter Depth (cm) - 0.3299 ± 0.4144 2.53 a 0.11
Distance to Water (m) - 0.0155 ± 0.0206 2.2600 0.13
Slope (degrees) 0.0709 ± 0.1076 1.74 a 0.19
Model 33.99 b < 0.0001
a Wald x2 b -2 Log Likelihood x2 c AIC = 34.68 30
Table 7. Results of multivariate matched-pairs logistic regression using variables from previous microhabitat and mesohabitat analyses to explain use of foraging locations by Varied Thrushes in Humboldt County, California, 1994-95. The analysis yielded two equivalent models. The first table depicts the most parsimonious model, based on the lower AIC (Akaike Information Criteria) value. Sample Size = 43.
2 Variable Coefficient ± 2 SE X P - value
Foliage Density (%) - 9.4952 ± 9.0406 4.41 a 0.036
Model c 49.14 b < 0.0001
2 a Wald x b -2 Log Likelihood x2
C AIC = 12.57
2 Variable Coefficient ± 2 SE X P - value
Microhabitat Shrub Cover (%) - 14.8789 ± 20.4294 2.12 a 0.14
Distance to Nearest Cover (m) - 2.2487 ± 3.5066 1.64 a 0.20
Mesohabitat Litter Depth (cm) - 0.3151 ± 0.4682 1.81 0.18
b Model c 52.39 < 0.0001 a Wald x2 b -2 Log Likelihood x2
C AIC = 13.83 Table 8. Standardized differences (± 2 SE) in standard deviations (SD) and mean differences (± 2 SE) in real units between bird- centered foraging location microhabitat (0.5 m circular plots) and mesohabitat (0.04 ha circular plots) attributes used by Varied Thrushes in Humboldt County, California, 1994-95. Sample Size = 43.
Variable SD ± 2 SE x Difference ± 2 SE t - value P - value
Forb Cover (%) 0.27 ± 0.23 7.94 ± 6.75 2.33 0.024 a
Bare Ground Cover (%) 0.17 ± 0.25 6.44 ± 6.92 1.36 0.18 a
Litter Cover (%) - 0.06 ± 0.20 - 3.99 ± 6.96 - 0.66 0.52 a
Litter Depth (cm) - 0.28 ± 0.23 - 1.03 ± 0.54 - 2.44 0.019 a
c b Shrub Cover (%) - 36.43 ± 6.03 - 12.09 < 0.0001
b Wood < 30 cm Cover (%) C 0.85 ± 1.85 0.92 0.36
b Wood > 30 cm Cover (%) - C - 3.85 ± 7.00 1.09 0.28 a Corresponds to one-sample t-test of standardized differences. b Corresponds to paired t-test of mean differences. c No standardized differences exist because the line-intercept method does not yield a distribution of values. 32 plot scale. Foraging microsites also had significantly greater forb cover and less litter depth than what was available within each bird-centered plot (Table 8).
I found a very strong association in the species of shrubs used by Varied Thrushes as foraging locations compared to random locations (Cramer's V = 1.00, n= 9). All
Varied Thrushes foraged in fruiting berry bushes while only two of the shrubs from the random locations were berry bushes which contained no berries at that time. Seven of nine birds foraged on red huckleberry (Vaccinium parvifolium), one on thimbleberry
(Rubus parviflorus), and one on salmonberry (Rubus spectabilis), whereas at random locations six of nine shrubs were sword fern (Polystichum munitum), one was hazelnut
(Corylus cornuta), one was red elderberry (Sambucus racemosa), and one was
California huckleberry (Vaccinium ovatum). I never observed foraging on berries before
1 June. DISCUSSION
Song Posts
In territorial, sexually dimorphic species such as the Varied Thrush (Dawson
1923, Bent 1949, Ehrlich et al. 1988), song and plumage advertise a male's fitness, both in terms of interactions with conspecific males, and for courtship and breeding activities with females. Not only are song quality (Morton 1975, Marten and Marler 1977, Shieck
1997) and plumage characteristics (Cott 1940, Endler 1978, Gotmark and Hohlfalt
1995) important, but the selection of a specific song post location has consequences as well (Marten and Marler 1977, Hunter 1980, Sedgwick and Knopf 1992, Weins 1989).
The characteristics of the song post will affect the transmission of both audio and visual information to conspecifics, in addition to potential predators and interspecific competitors (Hunter 1980). Marten and Marler (1977) found that acoustic attenuation of all frequencies was most pronounced less than 2 m above the ground in temperate habitats, including coniferous forests. Varied Thrushes solve the attenuation problem by singing in trees. However, any additional reduction in attenuation is negligible above
2 m (Marten and Marler 1977, Roberts et al. 1979). Thus there must be other reasons for using higher song posts. There are probably a greater number of song perches available within the canopy of trees, though the well developed, multi-layered subcanopy and understory of redwood forests provide song posts at all heights. Visual
33 34 detections of conspecifics and predators may be enhanced from higher perches. Song posts above the ground are generally more conspicuous, which might prove beneficial in interactions with conspecifics, though the risk from avian predators might increase as well (Hunter 1980).
Male Varied Thrushes sang from larger song post trees than those available both within bird-centered locations and between bird-centered and random locations within territories. Depending on the position of the song post within the tree, large, tall trees may provide more conspicuous song posts than smaller trees. Or the use of larger trees may simply reflect the higher probability of landing on a tree with a greater surface area and more perches. The greater tendency to use coniferous trees reflects the larger size of most conifers compared to deciduous trees. Song post trees were surrounded by a higher density of trees compared to random locations. This would account for the greater total foliage density found within bird-centered plots and make song posts less visible, especially over greater distances. Song posts were on steeper slopes than random locations, though a three degree difference may not be biologically meaningful.
Finally, song posts were closer to water, possibly because territories were centered in drainages.
Many song post characteristics, such as perches with low foliage density that were located close to the top of a larger tree within the bird-centered plot, are consistent with the perspective that male Varied Thrushes select conspicuous song posts. Since Varied
Thrushes are sexually dimorphic and males have brighter plumages, this seems reasonable. However, many song posts were within 20 m of the ground. Here, low 35 foliage density may increase sound transmission as well as visibility, or my sample of conspicuous song posts may simply have been biased toward visible birds. Also, any advantage conferred by a conspicuous song post would tend to be countered by the low visibility of closed-canopy redwood forests. Finally, conspicuous song posts within forest habitat might increase risk from avian predators compared to more open habitats where avian predators are more easily detected from a distance (Hunter 1980). In addition, a bright plumage does not necessarily confer high visibility (Cott 1940, Endler
1978, Gotmark 1995, Gotmark and Hohlfalt 1995). Varied Thrushes are similar to other forest dwelling birds (Gotmark 1995) in that they have a high-contrast, cryptic plumage that tends to camouflage and conceal them over longer distances amid the mosaic of light and dark patterns found in mature forests (Endler 1993). At close quarters a male's bright plumage would still be very conspicuous and advantageous during interactions with conspecifics, while at longer distances the song would more effectively advertise fitness without exposing males to increased risk of avian predation. Singing males may also conceal their locations with their song as well. As Dawson (1923) wrote, " He flings his voice at you as well from the tree-top as from the ground, now right, now left, the while he sits motionless upon a branch not fifteen feet above you."
Foraging Habitat
The tendency of animals to forage in well-defined microhabitats has been well documented (MacArthur 1958, Gibb 1964, Hutto 1981, Brush and Stiles 1986) and can be considered a logical extension of patch choice in foraging theory (Morse 1980, Hutto
1985, Brown and Morgan 1995). Breeding Varied Thrushes often foraged on the 36 ground for invertebrates (pers. obs.) in disturbed areas (especially along trails) within the forest. Ground foraging locations had less shrub cover, foliage density, woody debris < 30 cm diameter cover, litter cover, litter depth, and more bareground and forb cover (mostly grass in parks and yards) at both the mesohabitat and microhabitat scale.
However, less foliage density at the microhabitat scale best characterized ground foraging locations (Table 7). This was further supported by microsites having less shrub cover and greater distance to nearest cover, both of which were correlated with foliage density, and are evidence that little ground level vegetation was present at these sites.
As an alternative explanation, Varied Thrushes appeared to forage in open areas because those birds were more visible.
Varied Thrushes foraged closer to water, although, unexpectedly, foraging locations had less soil moisture at the microsite level. This apparent contradiction can best be explained by trails or other disturbed areas having less foliage density or cover of any kind (shrub and litter). The lack of cover would result in moisture loss from the soil through increased evaporation. Ground foraging locations were not necessarily dry, only drier compared to shaded, random locations generally located off trails within undisturbed forest. If moister sites were indeed important in providing a greater number of invertebrate prey, this would not be expected. Therefore, distance to water was probably more closely tied to territories centered in drainages.
The use of specific foraging locations (patch choice) is strongly affected by diet choice (Brown and Morgan 1995) based on the availability of temporally changing food resources and nutritional requirements. Though I did not measure food availability and 37 choice, I did observe changes in patch choice, similar to those that Hutto (1981) found in western wood warblers. This was evidence that Varied Thrushes closely tracked food resources through time. Early in the breeding season, when protein requirements of females and nestlings were presumably high, Varied Thrushes foraged predominantly on the ground for invertebrates (pers. obs.). After the young fledged, thrushes began including fruit in their diet; this was primarily red huckleberries which coincidentally became available in large numbers at that time (1 June). Though my sample size was small, the use of berries in the diet is supported by others (Dawson 1923, Bent 1949,
Ehrlich et al. 1988). Thus, in addition to habitat physiognomy, floristic composition
(that is red huckleberry) became an important explanatory variable of foraging locations as well (Rotenberry 1985).
Scale Perspectives
My results are evidence for the efficacy of using a hierarchical approach, both in terms of measuring habitat characteristics at the correct scale of resolution and determining the scale or scales at which animals select habitat for specific activities.
Microhabitat selection can be viewed as a hierarchical, scale-dependent process by which an animal chooses locations within a habitat (Hutto 1985). The varying spatial scales at which plumage, audio characteristics of song, and song post location operate are evidence that multiple scales were involved in the selection of song posts by Varied
Thrushes. In support of this, variables measured at microhabitat and mesohabitat scales were included in the final multivariate logistic regression model and were significant in the univariate analyses as well. At the territory size scale (Sedgwick and Knopf 1992), 38 singing males used song posts on steeper slopes that were closer to water compared to random locations. This provides evidence that territories appeared to be centered within steep drainages. At the within stand or mesohabitat scale, male Varied Thrushes chose sites with higher tree density, while at the microhabitat scale song post tree size, position within the tree, and foliage density around the perch site were important.
Although shrub cover was significant at both the mesohabitat and microhabitat scales, less microhabitat shrub cover, either within bird-centered or between bird- centered and random locations, was far more important than mesohabitat shrub cover.
This highlights the importance of measuring attributes at the correct spatial scale and corroborates earlier studies. Petit et al. (1988) found that mesoplot (0.04-ha circular plot) scale measurements diluted actual values of specific nest-site characteristics. In addition, mesoplots failed to capture the fine scale response to habitat heterogeneity by
Hooded Warblers (Wilsonia citrina) and Wood Thrushes (Hylocichla mustelina) in selecting nest-site locations; within 1 m radius of Hooded Warbler nests and 5 m radius of Wood Thrush nests. The 0.5 m circular plot used in measuring microhabitat characteristics usually fell within the relatively homogenous habitat of disturbed areas such as long, narrow trails, and therefore provided the correct scale of resolution in describing such foraging locations. However, 0.04 ha circular mesoplots sampled not only the trail, but much of the surrounding forest vegetation, thus incorporating a great deal of heterogeneity into the sample. In the future, researchers should not only determine the correct spatial scale, but the correct spatial configuration of the sampling unit in their design as well. As an added note, the ground foraging models were 39 probably superior to the song post models because I was able to more precisely measure attributes of easily accessible foraging locations on the ground compared to song posts high in trees.
The importance of foliage density, as well as the other microhabitat variables shrub cover and distance to nearest cover, also provided evidence that Varied Thrushes selected foraging sites predominantly at the microhabitat scale. Nonetheless, like song post locations, Varied Thrushes did select foraging sites at multiple scales and spatial configurations as well. Hilden (1965) and Hutto (1985) suggested that at larger scales, birds were more likely to respond to proximate cues, such as vegetation composition and structure, in assessing habitat suitability, whereas microsite use within a habitat was often based on active exploration in response to cues that were either very closely correlated with ultimate factors directly related to fitness, such as food or predation risk, or were the ultimate factors themselves. When foraging for red huckleberries, the bright red berries were both a proximate cue and the ultimate factor itself In selecting ground foraging locations, trails or other areas with little cover provided proximate cues for patch choice. Once a thrush settled within such an area, it then began actively searching for food items. These areas were probably strongly correlated with presence of invertebrates which would be more visible than in heavily vegetated surrounding locations. Areas with little ground level vegetation would also increase the detectability of terrestrial predators and reduce the risk of predation (Bouskila 1995). In this way the birds probably used a hierarchical process in selecting foraging locations. This is consistent with the hierarchical model of habitat selection proposed by Hutto (1985) as 40 well as signifying the importance of measuring habitat characteristics at multiple spatial scales (Petit et al. 1988, Sedgwick and Knopf 1992). LITERATURE CITED
AKAIKE, H. 1973. Information theory and extension of the maximum likelihood principle, pp. 267-281. In B.N. Petram, and F. Csaaki [eds.], Second international symposium on information theory. Akadeemia Kiadi, Budapest, Hungary.
ANDERSON, B. W. AND R. D. OHMART. 1986. Vegetation, pp. 639-660. In A.Y. Cooperider, R.J. Boyd, and H.R. Stuart [eds.], Inventory and monitoring of wildlife habitat. U.S. Dept. Inter. Bur. Land Manage. Service Center, Denver, CO.
BELL, G. W., S. J. HEM, AND J. VERNER. 1990. Proportional use of substrates by foraging birds: model considerations of first sightings and subsequent observations. Stud. Avian Biol. 13:161-165.
BENT, A. C. 1949. Life histories of North American thrushes, kinglets, and their allies. Dover Publications, Inc., New York, NY.
BONHAM, C. D. 1989. Measurements for terrestrial vegetation. John Wiley and Sons, Inc, New York, NY.
BOUSKILA, A. 1995. Interaction between predation risk and competition: a field study of kangaroo rats and snakes. Ecology 76:165-178.
BROWN, E. R. 1985. Management of wildlife and fish habitats in forests of western Oregon and Washington, part 2 - appendices. U. S. Dept. of Agric., Forest Service, Pacific Northwest Region, Portland, OR.
BROWN, J. S., AND R. A. MORGAN. 1995. Effects of foraging behavior and spatial scale on diet selectivity: a test with fox squirrels. Oikos 74:122-136.
BRUSH, T., and E. W. STILES. 1986. Using food abundance to predict habitat use by birds, pp. 57-63. In J. Verner, M.L. Morrison, and C.J. Ralph, [eds.], Wildlife 2000: modeling habitat relationships of terrestrial vertebrates. Univ. of Wisconsin Press, Madison, WI.
COLLINS, B. M., AND F. M. WHITE. 1981. Elementary forestry. Reston Pub. Co., Reston, VA.
41 42
COLLINS, S. L. 1981. A comparison of nest-site and perch-site vegetation structure for seven species of warblers. Wilson Bull. 93:542-547.
Corr, H. B. 1940. Adaptive coloration in animals. Methuen, London, UK.
DAWSON, W. L. 1923. The birds of California. Vol. 2. South Moulton Co., Los Angeles, CA.
DUNK, J. R., R. N. SMITH, AND S. L. CAIN. 1997. Nest-site selection and reproductive success in Common Ravens. Condor 114:116-120.
EHRLICH, P., D. S. DOBKIN, AND D. WHEYE. 1988. The birder's handbook. Simon and Shuster, Inc., New York, NY.
ENDLER, J. A. 1978. A predator's view of animal colour patterns. Evol. Biol. 11:319- 364.
ENDLER, J. A. 1993. The color of light in forests and its implications. Ecol. Monogr. 63:1-27.
GIBB, J. 1964. Feeding ecology of tits, with notes on treecreeper and goldcrest. Ibis 96:513-543.
GOTMARK, F. 1995. Black-and-white plumage in male Pied Flycatchers (ficedula hypoleuca) reduces the risk of predation from Sparrowhawks (Accipiter nisus) during the breeding season. Behavioral Ecol. 6:22-25.
GOTMARK, F., AND A. HOHLFALT. 1995. Bright male plumage and predation risk in passerine birds: are males easier to detect than females? Oikos 74:475-484.
GREENBERG, R. 1990. Ecological plasticity, neophobia, and resource use in birds. Stud. Avian Biol. 13:431-437.
HEJL, S. J., J. VERNER, AND G. W. BELL. 1990. Sequential versus initial observations in studies of avian foraging. Stud. Avian Biol. 13:166-173.
HILDEN, 0. 1965. Habitat selection in birds: a review. Annales Zoologic. Fenn. 2:53- 75.
HINTZE, J. L. 1996. Number Cruncher Statistical System (NCSS), version 6.0.21., Kaysville, UT.
HOLWAY, D. A. 1991. Nest-site selection and the importance of nest concealment in the Black-Throated Blue Warbler. Condor 93:575-581. 43
HOSMER, D. W., AND S. LEMESHOW. 1989. Applied logistic regression. John Wiley and Sons, New York, NY.
HUNTER, M. L., JR. 1980. Microhabitat selection for singing and other behavior in Great Tits, Parus major: some visual and acoustical considerations. Anim. Behay. 28:468-475.
HURT, M. C. 1996. The influence of forest fragmentation and vegetation structure on the breeding distribution of Varied Thrushes in redwood forests. M.S. Thesis, Humboldt State University, Arcata, CA.
HUTTO, R. L. 1981. Seasonal variation in the foraging behavior of some migratory western wood warblers. Auk 98:765-777.
HUTTO, R. L. 1985. Habitat selection by nonbreeding, migratory land birds, pp. 455-473. In M.L. Cody [ed.], Habitat selection in birds. Academic Press, New York, NY.
JAMES, F. C., AND C. E. McCULLOCH. 1990. Multivariate analysis in ecology and systematics: panacea or pandora's box. Ann. Rev. Ecol. Syst. 21:129-166.
LACK, D. L. 1968. Ecological adaptations for breeding in birds. Methuen, London, UK.
LARSON, D. L., AND C. E. BOCK. 1986. Determining avian habitat preference by bird-centered vegetation sampling pp. 37-44. In J. Verner, M.L. Morrison, and C.J. Ralph [eds.], Wildlife 2000: modeling habitat relationships of terrestrial vertebrates. Univ. of Wisconsin Press, Madison, WI.
LEBRETON, J. D., K. P. BURNHAM, J. CLOBERT, AND D. R. ANDERSON. 1992. Modeling survival and testing biological hypotheses using marked animals: a unified approach with case studies. Ecol. Monogr. 62:67-118.
MACARTHUR, R. H. 1958. Population ecology of some warblers of northeastern coniferous forests. Ecology 42:599-619.
MARTEN, K., AND P. MARLER. 1977. Sound transmission and its significance for animal vocalization: I. Temperate forest habitats. Behay. Ecol. Sociobiol. 2:271- 291.
MARTIN, T. E., AND J. J. ROPER. 1988. Nest predation and nest-site selection of western population of the Hermit Thrush. Condor 90:51-57. 44
MOERMOND, T. C. 1988. A functional approach to foraging: morphology, behavior, and the capacity to exploit. Stud. Avian Biol. 13:427-430.
MORSE, D. H. 1980. Population limitation: breeding or wintering grounds?, pp. 505- 516. In A. Keast and E.S. Morton [eds.], Migrant birds in the Neotropics: ecology, behavior, distribution, and conservation. Smithsonian Institution Press, Washington, D. C.
MORTON, E. S. 1975. Ecological sources of selection on avian sound. Amer. Natur. 108:17-34.
NOON, B. R. 1981. Techniques for sampling avian habitat, pp. 42-52. In Capen, D.E. [ed.], The use of multivariate statistics in studies of wildlife habitat. U.S. Forest Serv. Gen. Tech. Rep. RM-87, U.S. Dept. of Agric., Forest Service, Rocky Mountain Forest and Range Experiment Station, Fort Collins, CO.
PETIT, K. E., D. R. PETIT, AND L. J. PETIT. 1988. On measuring vegetation characteristics of bird territories: nest sites vs. perch sites and the effect of plot size. Am. Midl. Nat. 119:209-215.
RECHER, H. F. 1988. Specialist or generalist: avian response to spatial and temporal changes in resources. Stud. Avian Biol. 13:333-336.
ROBERTS, J., A. KACELNIK, AND M. L. HUNTER JR. 1979. A model of sound interference in relation to acoustic communication. Anim. Behay. 27:1271-1273.
ROTENBERRY, J. T. 1985. The role of habitat in avian community composition: physiognomy or floristics? Oecologia 67:213-217.
SAKAI, H. F., AND B. R. NOON. 1991. Nest-site characteristics of Hammond's and Pacific-slope Flycatchers in northwestern California. Condor 93:563-574.
SAS INSTITUTE, INC. 1995. SAS proprietary software release, version 6.11. SAS Institute, Inc., Cary, NC.
SAWYER, J. 0., JR., AND T. KEELER-WOLF. 1995. A manual of California vegetation. California Native Plant Society, Sacramento, CA.
SEDGWICK, J. A., AND F.L. KNOPF. 1992. Describing Willow Flycatcher habitats: scale perspectives and gender differences. Condor 94:720-733.
SHIECK, J. 1997. Biased detections of bird vocalizations affects comparisons of bird abundance among forested habitats. Condor 99:179-190. 45
WEINS, J. A. 1989. The ecology of bird communities. Vol. 1. Foundations and patterns. Cambridge Univ. Press, New York, NY. PERSONAL COMMUNICATIONS
GOLIGHTLY, R. T. 1994. Department of Wildlife, Humboldt State University, Arcata, CA 95521.
SAWYER, J. 0., JR. 1994. Department of Biological Sciences, Humboldt State University, Arcata, CA 95521.
46 Appendix A. Common shrubs, ferns, and forbs of redwood forests in Northwestern California.
Common Name Scientific Name
SHRUBS
Vine maple Acer circinatum Oregon grape Berberis aquifolium Hazelnut Corylus cornuta Salal Gaultheria shallon Cascara Rhamnus purshiana California rose-bay Rhododendron macrophyllum Western azalea Rhododendron occidentale Red flowering currant Ribes sanguineum Wood rose Rosa gymnocarpa Himalayan blackberry Rubus discolor Thimbleberry Rubus parviflorus Salmonberry Rubus spectabilis California blackberry Rubus ursinus Red elderberry Sambucus racemosa Poison oak Toxicodendron divers ilobum California huckleberry Vaccinium ovatum Red huckleberry Vaccinium parvifolium
FERNS
Five-finger fern Adiantum aleuticum Lady fern Athyrium filix-femina Deer fern Blechnun spicant Wood fern Dryopteris expansa Western sword fern Polystichum munitum California polypody Polypodium californicum Bracken fern Pteridium aquilinum Giant chain fern Woodwardia fimbriata
47 48
Appendix A. Common shrubs, ferns, and forbs of redwood forests in Northwestern California (continued).
Common Name Scientific Name
FORBS
Yarrow Achillea millefolium Vanilla leaf Achlys triphylla Trail plant Adenocaulon bicolor Pearly everlasting Anaphalis margaritacea Windflower Anemone deltoidea Wild ginger Asarum caudatum Sedges Carex spp. Miner's lettuce Claytonia perfoliata Red clintonia Clintonia andrewsiana Pampas grass Cortaderia selloana Queen Anne's lace Daucus carota Fairy bells Disporum hookeri Horsetail Equisetum spp. Bedstraw Galium spp. Douglas iris Iris douglasiana Rushes Juncus spp. Honeysuckle Lonicera hispidula Yellow skunk cabbage Lysichiton americanum Redwood sorel Oxalis oregana Grasses Poaceae Western coltsfoot Petasites frigidus Selfheal Prunella vulgaris Curly dock Rumex crispus California bee-plant Scrophularia californica Fat Solomon's seal Smilacina racemosa Thin Solomon's seal Smilacina stellata Hedge nettle Stachys ajugoides Sugar scoops Tiarella unifoliata Fringe cups Tellima grandiflora Pig-a-back plant Tolmiea menziessii Pacific star flower Trientalis latifolia Western wake robin Trillium ovatum White inside-out flower Vancouveria hexandra Evergreen violet Viola sempervirens Yerba de selva Whipplea modesta