Perch Trees and Shoreline Development as Predictors of Bald Eagle Distribution on Chesapeake Bay

Abstract

We studied the influence of shoreline perch trees and human development on bald eagle (Haliaeetus leucocephalus) distribution on the northern Chesapeake Bay. Bald eagle distributions may be determined by available suitable shoreline perch areas. Models based on human development and shoreline habitat variables may alleviate problems associated with classifying bald eagle habitat by identifying characteristics predictive of eagle presence. We observed 2,962 eagles during 36 shoreline surveys and relocated 110 radio-marked eagles 1,350 times during 1985-92. We counted 5,928 suitable (height ≥6.1 m, diam at breast height [dbh] ≥20.0 cm, and shoreline accessibility ≥30°) perch trees in 229, 250- × 50-m segments along shoreline during 1990-91. Shoreline segments used by eagles had more suitable perch trees (x̄ = 30.3 vs. 22.0; P < 0.001) and a larger percentage of forest cover (x̄ = 54.9 vs. 39.4; P < 0.001) than unused segments. Suitable trees on segments with eagle use were closer to water than suitable trees on segments without eagle use (x̄ = 8.4 vs. 17.0 m; P = 0.009). Most segments classified as marsh (66.7%) were unused. Marsh segments had fewer suitable perch trees, less forest cover, and a greater mean distance from water to the nearest suitable perch tree than did other land types (P < 0.001). Developed segments had fewer suitable perch trees, less forest cover, and a shorter distance from water to the nearest suitable perch tree than undeveloped forested segments (P ≤ 0.01). Logistic regression models based on various measures of perch tree abundance and shoreline development correctly predicted eagle use for 65.9-71.0% of segments.

Full text

Wiley and Wildlife Society
are collaborating with JSTOR to digitize, preserve and extend access to
The Journal of Wildlife
Management.
http://www.jstor.org
Perch Trees and Shoreline Development as Predictors of Bald Eagle Distribution on Chesapeake
Bay
Author(s): Sheri K. Chandler, James D. Fraser, David A. Buehler and Janis K. D. Seegar
Source:
The Journal of Wildlife Management,
Vol. 59, No. 2 (Apr., 1995), pp. 325-332
Published by: on behalf of the Wiley Wildlife Society
Stable URL: http://www.jstor.org/stable/3808946
Accessed: 05-01-2016 16:18 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
This content downloaded from 160.36.178.25 on Tue, 05 Jan 2016 16:18:33 UTC
All use subject to JSTOR Terms and Conditions

PERCH
TREES
AND
SHORELINE DEVELOPMENT AS PREDICTORS
OF BALD EAGLE
DISTRIBUTION
ON CHESAPEAKE
BAY
SHERI
K. CHANDLER,
Department
of Fisheries
and Wildlife
Sciences, Virginia
Polytechnic
Institute
and State University,
Blacksburg,
VA
24061-0321, USA
JAMES
D. FRASER,
Department
of Fisheries
and
Wildlife
Sciences, Virginia Polytechnic
Institute and State University,
Blacks-
burg,
VA
24061-0321, USA
DAVID A. BUEHLER,'
Department
of Fisheries and
Wildlife
Sciences, Virginia Polytechnic
Institute
and State University,
Blacks-
burg,
VA
24061-0321, USA
JANIS
K. D. SEEGAR,
Chemical
Research,
Development,
and Engineering
Center,
U.S. Army,
Aberdeen
Proving
Ground,
MD
21010-5423, USA
Abstract: We studied the influence of shoreline
perch trees and human development on bald eagle (Hal-
iaeetus leucocephalus)
distribution
on the northern
Chesapeake Bay. Bald eagle distributions
may be deter-
mined by available
suitable shoreline
perch
areas.
Models
based
on human
development
and shoreline habitat
variables
may alleviate problems
associated
with classifying
bald eagle habitat
by identifying characteristics
predictive
of eagle presence.
We observed
2,962 eagles during
36 shoreline
surveys
and relocated
110 radio-
marked
eagles 1,350 times during 1985-92. We counted
5,928 suitable
(height >6.1 m, diam at breast
height
[dbh] _20.0 cm, and shoreline
accessibility
>300) perch trees in 229, 250- x 50-m segments
along shoreline
during 1990-91. Shoreline
segments used by eagles had more suitable perch trees (f = 30.3 vs. 22.0; P <
0.001) and a larger
percentage
of forest cover (: = 54.9 vs. 39.4; P < 0.001) than unused
segments.
Suitable
trees on segments with eagle use were closer to water than suitable trees on segments
without eagle use (i
= 8.4 vs. 17.0 m; P = 0.009). Most segments classified as marsh
(66.7%)
were unused. Marsh
segments had
fewer suitable
perch trees, less forest cover, and a greater
mean distance
from water to the nearest
suitable
perch tree than did other land types (P < 0.001). Developed segments had fewer suitable perch trees, less
forest cover, and a shorter distance from water to the nearest suitable
perch tree than undeveloped
forested
segments (P < 0.01). Logistic regression
models based on various measures
of perch tree abundance and
shoreline
development correctly
predicted eagle use for 65.9-71.0%
of segments.
J. WILDL.
MANAGE.
59(2):325-332
Key words: bald eagle, Chesapeake
Bay, development, habitat, Haliaeetus leucocephalus,
human distur-
bance, logistic regression, Maryland.
A key task of bald eagle recovery plans is to
identify and monitor eagle habitat (Grier et al.
1983, Murphy et al. 1989, Byrd et al. 1990).
Much of this work has been done by surveying
shoreline habitat and noting where eagles occur.
Whereas this method often has been productive,
suitable areas may be overlooked because of the
large amount of shoreline to be surveyed and
because eagle populations are below carrying
capacity, leaving some suitable habitat vacant.
These problems can be circumvented by iden-
tifying habitat characteristics predictive of ea-
gle presence, and then surveying habitats for
these characteristics.
Many studies have shown that bald eagles
avoid areas frequented by people (Stalmaster
and Newman 1978, Knight and Knight 1984,
Fraser et al. 1985, Buehler et al. 1991, Mc-
Garigal et al. 1991). Although it has been sug-
gested that perching habitat may determine ea-
gle distribution and abundance (Stalmaster and
Newman 1979, Fielder and Starkey 1980, Steen-
hof et al. 1980), this hypothesis has not been
tested. Moreover, human development and hab-
itat variables have not been used together to
predict eagle use of shoreline. Application of
univariate models, when several variables affect
a response, can lead to spurious results (Fienburg
1980).
We tested the hypothesis that the probability
of bald eagle use of shoreline increases with the
number of suitable perch trees near the shore.
We also examined the hypothesis that eagle
avoidance of developed areas is related to scar-
city of suitable perch trees. We used measures
of perch tree abundance and shoreline devel-
opment to create logistic regression models that
I Present address:
Department of Forestry, Wild-
life and Fisheries,
P.O. Box 1071, University
of Ten-
nessee, Knoxville,
TN 37901, USA.
325
This content downloaded from 160.36.178.25 on Tue, 05 Jan 2016 16:18:33 UTC
All use subject to JSTOR Terms and Conditions

326 BALD
EAGLE
SHORELINE PERCH USE * Chandler et al. J. Wildl. Manage. 59(2):1995
enable objective assessments of the suitability of
shorelines for perching eagles.
Research was funded by the U.S. Army
Chemical Research, Development and Engi-
neering Center, Aberdeen Proving Grounds. We
thank J. P. Ondeck, W. S. Seegar, and F. P.
Ward for logistical support, and E. M. Adams,
P. L. Andrews, L. L. Arnold, W. E. Buford, B.
A. Buehler, C. Campbell, L. A. Colby, A. M.
Davidson, A. K. DeLong, D. C. DeLong, Jr., S.
A. Eisner, K. W. Hegstad, W. M. Iko, D. W.
Liedlich, K. F. McCabe, J. L. McConnaughey,
T. J. Mersmann, M. B. Moss, M. C. Ogle, M.
Roeder, J. M. Seegar, L. A. Serlin, and T. L.
Weller for help with data collection and/or
computerization. A. B. Jones III and D. J. Dean
assisted with ARC/INFO analyses. We thank
our pilot, J. V. Kesser. M. P. Caussey, K. W.
Cline, C. A. Koppie, T. A. Steffer climbed nest
trees and helped radiomark nestlings. We thank
landowners for allowing us access to shorelines.
We thank R. L. Kirkpatrick and D. F. Stauffer
for comments on the manuscript.
STUDY
AREA
We conducted research along 580 km of the
northern Chesapeake Bay shoreline in Mary-
land. Water near shore was shallow; depths 20
m from mean high tide line varied from 0.2 to
1.2 m. Shoreline slope within 20 m of the mean
high tide line varied from 5 to 60*. Cliffs 6-25
m high were common on the eastern shore of
the Bay. Marshes occurred along rivers and creeks
throughout the study area but were largest on the
western shoreline. The most common tree group
was oak (Quercus spp.). Sweetgum (Liquidambar
styraciflua) and ash (Fraxinus spp.) were locally
abundant, and black cherry (Prunus serotina),
black locust (Robinia pseudoacacia), elm (Ulmus
spp.), and tuliptree (Liriodendron tulipifera) were
present (Harlow et al. 1991).
Buildings and other developments were com-
mon and irregularly distributed along the shore-
line. Pedestrian and boating activity were com-
mon along shoreline areas and frequent near
public parks, housing developments, and ma-
rinas.
METHODS
Capture and Radiomarking
Using floating noose-fish (Cain and Hedges
1989) or padded leghold traps (Young 1983), we
captured adult bald eagles. We radiomarked 32
adult bald eagles from 1984 to 1988 and 78
eaglets 8-10 weeks old during 1984-91, fitting
them with 65-g radio transmitters equipped with
solar-charged, nickel-cadmium batteries. We
mounted transmitters dorsally on each eagle in
a back-pack configuration and attached them
with 1-cm-wide, dark brown, teflon ribbon
(Buehler et al. 1991).
Telemetry
and Shoreline
Eagle Surveys
We conducted telemetry flights twice weekly
during 1985-88, and twice monthly from 1989
to 1992 in a fixed-wing aircraft. We flew 80-
130 km/hour at altitudes between 30 and 300
m. When > 10 radio-marked birds were on the
study area, we randomly selected 10 for relo-
cation to minimize fatigue.
We conducted shoreline surveys monthly in
fixed-wing aircraft from September 1985 to Au-
gust 1988 and September 1991 to May 1992.
We flew at 80-110 km/hour at approximately
30-50 m altitude. We flew over water, beach,
or marsh passing within 50 m of the trees closest
to water. Surveys began 30 minutes after sunrise
and ended within 3.5 hours after sunrise. We
visited all study area shoreline in each survey.
We recorded locations of 3,887 perched eagle
sightings from survey flights and radio-marked
eagles on 7.5-minute topographic maps and re-
corded Universal Transverse Mercator coordi-
nates of each location to the nearest 10 m.
Characterizing
Shoreline
Segments
We divided the shoreline into 2,340 segments
using a digitized version of the 7.5-minute to-
pographic maps. We considered a segment used
if ?1 eagle was perched there. If we recorded
no eagles on a segment, we considered it unused.
We randomly selected 103 of 742 used segments
and 126 of 1,598 unused segments for further
investigation. All selected segments measured
250 m along the shore (the i eagle flush distance
in east. North Am.; Stalmaster and Newman
1978, Wallin and Byrd 1984, Smith 1988, Bueh-
ler et al. 1991) and 50 m in width. We randomly
selected an additional 100 segments from the
remaining surveyed segments for verifying the
final logistic regression model. We noted eagle
use of these segments after selection: 33 were
used and 67 were unused.
Eagles use a variety of tree species for perch-
ing, provided the tree is large with an open,
spreading form and stout, horizontal limbs (Stal-
master 1987). Hence, we considered a tree suit-
This content downloaded from 160.36.178.25 on Tue, 05 Jan 2016 16:18:33 UTC
All use subject to JSTOR Terms and Conditions

J. Wildl. Manage. 59(2):1995 BALD EAGLE SHORELINE PERCH
USE * Chandler et al. 327
able for perching if it was at least as large as
the smallest perch tree observed by Buehler et
al. (1992) (height >6.1 m and dbh 220.0 cm)
and had >30* accessibility from the shoreline.
For each suitable perch tree, we recorded height
above mean high tide line (shoreline height).
Because eagles on the Chesapeake Bay prefer
dead to live trees for perching (Buehler et al.
1992), we noted tree condition (live or dead).
For each segment, we measured horizontal dis-
tance from mean high tide line to the closest
tree and calculated mean shoreline height and
the range of shoreline heights.
We classified each segment as forest, marsh,
agricultural, developed, or other, on the basis of
land cover type occupying the largest propor-
tion of the segment. We estimated percent forest
cover for each segment by outlining the segment
on a 1:12,500 color aerial photo. We laid a 1-
x 1-mm dot grid on the photo and calculated
the percentage of dots occurring on trees for
each segment. For segments with <50% forest,
we ocularly determined land type covering the
largest proportion of the segment. When land
type could not be determined ocularly, we used
the dot grid method to estimate dominant land
cover type. We considered a segment developed
if it contained a building. Segments categorized
as other included 4 Aberdeen Proving Ground
munitions test ranges, 3 lawns, 2 beaches, and 1
fencerow. We used Advanced Revelation com-
puter software (Revelation Tech., Inc. 1991) to
determine distance from each segment midpoint
to the closest building within 1,000 m and density
of buildings within 500 m of the segment mid-
point.
Analyses
Because all data were nonnormal, we used
nonparametric Wilcoxon rank sum and Kruskal-
Wallis tests for differences among distribution
medians. We used the Chi-square test of equal
proportions to determine if use of land types
was different from expected.
We used logistic regression models to predict
probability of eagle use (eagle use index) of a
shoreline segment given study conditions and
methods. If the predicted probability of eagle
use of a segment was ?0.5, we considered the
segment used for modeling purposes. We did
not use observations with incomplete data for
-1 variable in model development, thus cre-
ating unequal sample sizes among models. We
used stepwise variable selection to develop mod-
Table 1. Mean number
of suitable
(height -6.1 m, diam at
breast
height
>
20.0 cm, and _
30"
accessibility
from the shore-
line)
live trees and suitable
dead trees, mean percent
forest
cover, and mean distance
from water
to closest suitable tree
(m) on 250- x 50-m shoreline
segments, classified by bald
eagle use, on northern
Chesapeake
Bay, Maryland,
1990-92.
Characteristic na f SE Range pb
Suitable
trees
Eagle use 103 30.3 1.82 0-67.0
No eagle use 126 22.0 1.67 0-69.0 <0.001
Dead trees
Eagle use 103 1.2 0.14 0-7.0
No eagle use 126 0.8 0.12 0-8.0 0.019
Forest cover
Eagle use 102 54.9 3.39 0-100
No eagle use 125 39.4 3.18 0-100 <0.001
Distance to water
Eagle use 100 8.3 1.57 0-50.1
No eagle use 122 16.8 1.94 0-50.1 0.009
a Missing data for some segments resulted in unequal sample sizes
among characteristics.
b Wilcoxon rank sum test of Ho: values between use categories are
equal.
els. We investigated effects of 8 variables on
eagle use: number of suitable perch trees, per-
cent forest cover, distance from water to nearest
suitable perch tree, average shoreline height,
range in shoreline height, number of snags, de-
velopment density, and distance from shore to
nearest development. We calculated Cohen's
Kappa (K) statistic as a measure of the model's
ability to predict eagle use better than chance
(Titus et al. 1984). A K > 0 indicates fewer
misclassifications than if segments were classi-
fied by chance alone. We tested the null hy-
pothesis that K = 0 using a Z-test for each model.
We validated the final logistic regression equa-
tion using 100 randomly selected shoreline seg-
ments that were not used to create original equa-
tions.
RESULTS
Eagle Use of Shoreline
Segments
We identified 5,928 suitable perch trees on
57.3-km shoreline studied (1 tree/10 m). Ninety
percent of perched eagles observed (3,481 of
3,887) were in trees. Shoreline segments with
eagle use had more suitable perch trees, more
suitable dead trees, and greater percent forest
cover than did segments that were not used (Ta-
ble 1). Distance from open water to nearest suit-
able perch tree was shorter for used segments
than for unused segments. Of radio-marked ea-
gles perched within 50 m of the shoreline, 84.9%
were within 10 m.
This content downloaded from 160.36.178.25 on Tue, 05 Jan 2016 16:18:33 UTC
All use subject to JSTOR Terms and Conditions

328 BALD
EAGLE
SHORELINE
PERCH USE * Chandler et al. J. Wildl. Manage. 59(2):1995
Table
2. Land
types of segments with and without recorded bald
eagle use on northern
Chesapeake Bay, Maryland,
1990-
92.
Land typea
Forest Developed Marsh Agricultural Otherb
Eagle use n % n % n % n % n %
Use
Observed 51 60.0 11 26.2 26 33.8 8 57.1 7 63.6
Expected 38 44.7 19 45.1 35 45.5 6 42.9 5 44.5
Residuals! 2.3 -2.8 -1.7 2.2 2.9
No use
Observed 34 40.0 31 73.8 51 66.2 6 42.9 4 36.4
Expected 47 55.3 23 54.8 42 54.5 8 57.1 6 55.5
Residualse -2.1 2.6 1.6 -1.9 -2.6
Total 85 42 77 14 11
a Land types differed between used segments and segments without recorded eagle use, based on Chi-square test of equal proportions (x2 =
20.04, 4 df, P < 0.001).
b Included 4 Aberdeen Proving Ground munitions test ranges, 3 lawns, 2 beaches, and 1 fencerow.
0
Standardized residuals: (observed - expected)/square root of expected.
Eagles used forest segments more than ex-
pected and marsh and developed segments less
than expected (Table 2). The number of suitable
perch trees differed among land types (Table
3). Forested shoreline had more suitable perch
trees than developed, marsh, agricultural, and
other shoreline. Of 42 segments with no trees,
we classified 41 (97.6%) as marsh. Percent forest
cover also differed among land types (Table 3).
Marsh segments had the least amount of forest
cover, and developed, agricultural, and other
segments had less forest cover than forested seg-
ments. Likewise, the distance from open water
to nearest suitable tree differed among land
types. Distance from open water to nearest suit-
able perch tree was greater on marsh shoreline
than on developed, forest, other, or agricultural
land types (Table 3).
Most marsh shoreline segments (66.2%, 51 of
77) were not used by eagles. Used marsh seg-
ments had more trees, a larger percentage of
forest cover, and a larger distance from water
to the nearest tree than did unused marsh (Table
3).
Predicting
Probability
of Eagle
Shoreline Use
Development density and distance from wa-
ter to nearest tree were the best predictors of
eagle shoreline use (K = 0.383, SE = 0.067, Z
= 5.716, P < 0.001; Tables 4 and 5, Model 1;
Fig. 1). The logistic regression equation con-
taining these variables predicted eagle shoreline
use correctly for 68.9% of 222 segments. When
the distance from water to the nearest tree vari-
able was removed as a dependent variable, step-
wise selection produced a model that included
development density and percent forest cover
(68.7% correct, K = 0.361, SE = 0.068, Z = 5.309,
P < 0.001; Tables 4 and 5, Model 2; Fig. 1).
When we also removed percent forest cover, the
number of suitable perch trees per segment en-
tered the model (65.9% correct, K = 0.304, SE
= 0.067, Z = 4.537, P < 0.001; Tables 4 and 5,
Model 3; Fig. 1). None of the 3 measures of
perch tree availability (distance from water to
nearest tree, %
forest cover, and no. of suitable
perch trees/segment) was significant in a model
simultaneously, indicating that these measures
provided redundant information. No other vari-
able had predictive value (P > 0.05). When we
applied Model 2 to the validation dataset, eagle
use was correctly predicted for 71.0% of the
segments (Table 5).
DISCUSSION
Bald eagles on northern Chesapeake Bay used
shoreline that had more suitable perch trees,
more forest cover, and fewer buildings than un-
used areas. These results are consistent with the
hypothesis that suitable perch tree availability
and human development combine to affect ea-
gle use of shoreline habitat.
In addition, distance from the water to the
nearest suitable tree was shorter for used seg-
ments than unused segments. This emphasizes
the need to conserve perch trees near shoreline.
According to the model, in the absence of de-
velopment, trees should be <20 m from the
water for a 50% probability of eagle use of shore-
line. This is consistent with the observation that
bald eagles tend to perch <50 m from shore
This content downloaded from 160.36.178.25 on Tue, 05 Jan 2016 16:18:33 UTC
All use subject to JSTOR Terms and Conditions

Table 3. Number
of suitable
(height
26.1 m, diam
at breast height a20.0 cm, and ?300 accessibility
from
the shoreline)
trees, percent
forest cover, and distance from
open water to the
closest suitable
perch
tree by land
type and bald
eagle use on 250- x 50-m shoreline
segments on northern
Chesapeake Bay, Maryland,
1990-92.
Land type
Forest Developed Agricultural Othera Marsh
Characteristicbe n f SE Range n SE Range n f SE Range n SE Range n X SE Range
Suitable trees
Used segmentsd 51 42.9 1.61 22-67 11 33.1 4.43 5-54 8 27.3 3.71 14-41 7 10.7 3.61 0-25 26 10.8 2.48 0-43
Unused segments 34 42.6 1.93 14-69 31 24.1 2.22 2-48 6 25.3 7.64 3-50 4 32.0* 7.01 15-44 51 5.8* 1.34 0-42
Total 85 42.8A 42 26.5B 14 26.4B 11 18.5B 77 7.5C
Forest Cover
Used segmentsd 50 78.7 2.72 30-100 11 64.5 9.70 6-100 8 47.1 7.07 25-77 7 15.6 6.02 0-45 26 18.0 3.80 0-50
Unused segments 34 84.1 3.24 34-100 31 47.0 4.96 0-97 6 32.2 6.80 6-56 4 47.8 12.89 15-78 50 8.1* 1.97 0-53
Total 84 79.2D 42 51.6E 14 40.7EF 11 27.3F 76 11.5G
Distance to tree
Used segmentse 50 1.9 0.55 0-15.9 11 3.6 1.69 0-18.0 8 0.7 0.38 0-2.4 6 7.4 4.69 0-29.9 25 25.9 4.49 0-50.1
Unused segments 33 0.8 0.41 0-10.0 29 6.3 2.10 0-43.4 6 1.6 1.61 0-9.7 4 5.7 2.97 0-11.9 50 36.0* 2.85 0-50.1
Total 83 1.5H 40 5.51 14 1.5H 10 6.71 75 32.7J
a Included 4 Aberdeen Proving Ground munition test ranges, 3 lawns, 2 beaches, and 1 fencerow.
b Values for used and unused segments within columns followed by an asterisk (*) were different at P _ 0.05 using Wilcoxon rank sum test.
c Land types were different within characteristics (Kruskal-Wallis, 4 df, P < 0.001). Values followed by the same letter were not different (P : 0.01) based on Wilcoxon rank sum test.
d Used and unused segments differed overall (Wilcoxon rank sum, P < 0.001).
e Used and unused segments differed overall (Wilcoxon rank sum, P = 0.009).
C1
CA
e,
to
CD
Ul
tri
M
Cl
co
bo
co
This content downloaded from 160.36.178.25 on Tue, 05 Jan 2016 16:18:33 UTC
All use subject to JSTOR Terms and Conditions

330 BALD EAGLE
SHORELINE
PERCH
USE * Chandler et al. J. Wildl. Manage. 59(2):1995
0.8
0.6
0.4
0.2
0 5 10 15 20 25 30 35 40 45 50
DISTANCE FROM WATER TO CLOSEST TREE
0.8-
S0.6
0.4-
C
0o 10 20 30 40 50 60 70 80 90 100
% FOREST COVER
0.8
0.6
0.4
0.2
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
TREES/SEGMENT
+No buildings * 0.25 buildings/ha -0-0.5 buildings/ha
- 1.0 buildings/ha +*1.5 buildings/ha
Fig. 1. Effect
of distance
from water
to the nearest suitable
(height
>6.1 m, diam
at breast height _20.0 cm, and _30"
accessibility
from
the shoreline)
tree, percent
forest
cover,
and
number
of suitable
trees on the probability
of bald
eagle use
of shoreline
segments for different
development
densities
on
northern
Chesapeake
Bay, Maryland
1990-92. The
eagle use
index
is the probability
of observing _1 eagle, using
methods
described and under
conditions
existing at the time of this
study.
(Stalmaster and Newman 1979, Steenhof et al.
1980, Buehler et al. 1992). Proximity to water
probably provides eagles good visibility of the
water and a likelihood of capturing prey.
The overall difference between used and un-
used segments was largely influenced by marsh-
es that had (1) fewer trees, (2) a greater distance
from water to the nearest tree, and (3) lower
eagle use than the other habitats. The scarcity
of suitable perch trees and longer distance from
shore to perches probably deterred foraging ea-
gles.
Used segments also had more dead trees than
did unused segments. Although dead and dying
trees are preferred perch trees, most trees used
for perching on northern Chesapeake Bay are
live (Buehler et al. 1992). The larger number of
snags on used segments than unused areas may
simply reflect the larger number of trees on used
sections.
That bald eagles used developed shoreline less
than undeveloped shoreline is consistent with
their known avoidance of developed areas (Stal-
master and Newman 1978, Fraser et al. 1985,
Chester et al. 1990, Buehler et al. 1991, Mc-
Garigal et al. 1991). The few developed shore-
line segments that were used had more large
trees than did other segments. Our models pre-
dicted that as development density increased,
number of suitable perch trees and percent for-
est cover needed to retain the same probability
of use increased. Perhaps these trees provided
visual screening that prevented eagles from see-
ing people near developments, and therefore
kept eagles from being frightened away.
MANAGEMENT
IMPLICATIONS
The models we developed will help identify
bald eagle perching habitat on Chesapeake Bay
shoreline. Strictly interpreted, the eagle use in-
dex is an estimate of the probability that an eagle
would be observed on a segment under the con-
ditions (survey methods and frequency, eagle
population level, etc.) existing during this study.
Data needed for Models 1 and 3 are best
gathered in the field. In contrast, data needed
for Model 2 (development density and % forest
cover) can be obtained from maps, aerial pho-
tographs, or Geographic Information System.
Models correctly predicted bald eagle shore-
line use for 66-71% of the segments. Whereas
this level of accuracy can assist in land man-
agement for eagles, further improvement of
model accuracy is desirable. Because Chesa-
peake Bay shoreline property is valuable, know-
ing which areas should be preserved for eagles
and which can be developed without degrading
eagle habitat is important. The addition of vari-
This content downloaded from 160.36.178.25 on Tue, 05 Jan 2016 16:18:33 UTC
All use subject to JSTOR Terms and Conditions

J. Wildl. Manage. 59(2):1995 BALD EAGLE SHORELINE PERCH USE * Chandler et al. 331
Table
4. Logistic regressiona
parameter
estimates for
predicting
bald
eagle use of 250- x 50-m
shoreline
segments on northern
Chesapeake
Bay, Maryland,
1990-92.
Parameter estimates
Variable B SE X2 P
Model 1
Intercept 0.6857 0.2084 10.8286 0.001
Development densityb -2.5818 0.7086 13.2770 <0.001
Distance to tree (m) -0.0333 0.00809 17.0118 <0.001
Model 2
Intercept -0.5040 0.2392 4.4402 0.035
Development densityb -2.2878 0.6815 11.2707 <0.001
% forest cover! 0.0155 0.00413 13.9748 <0.001
Model 3
Intercept -0.4999 0.2446 4.1789 0.041
Development densityb -2.3177 0.6799 11.6193 <0.001
No. of perch trees 0.0280 0.0077 13.1664 <0.001
a The logistic regression equation was 0 = 1/(1 + exp[- {~o +
~Oj'xij}]), i = 1, 2. n, where 0 is the probability of eagle use,/ o is the beta
J-1
value of the intercept, Bj is the beta value of the j dependent variables, and xijs are data values for the k independent variables.
b Buildings/ha.
c % segment with forest cover.
Table
5. Logistic regression
equation
predictions
of bald
eagle
use of 250- x 50-m
shoreline
segments based on development
density
(buildings/ha),
distance
from
water
to the closest tree
(m), percent
forest
cover,
and number
of suitable
(height _6.1
m, diam at breast height 220.0 cm, and >30" accessibility
from
the shoreline)
trees on northern
Chesapeake Bay, Mary-
land,
1990-92.
Predicted use (segments) Total correct
Model
Observed use Eagle use No eagle use na %
Model 1
Development densityb, distance to tree (m)
Eagle use 75 25 100 75.0
No eagle use 44 78 122 63.9
Total (n) 119 103 222
Correct (%) 63.0 75.7 68.9
Model 2
Development densityb, % forest coverc
Eagle use 61 41 102 59.8
No eagle use 30 95 125 76.0
Total (n) 91 136 227
Correct (%) 67.0 69.8 68.7
Model 3
Development densityb, no. of perch trees
Eagle use 59 44 103 57.3
No eagle use 34 92 126 73.0
Total (n) 93 136 229
Correct (%) 63.4 67.6 65.9
Model 2 Validation
Eagle use 18 15 33 54.5
No eagle use 14 53 67 79.1
Total (n) 32 68 100
Correct (%) 56.3 78.0 71.0
a Missing data for some segments resulted in unequal sample sizes
among models.
b Buildings/ha.
% segment with forest cover.
ables such as the depth of the water adjacent to
the segment, or human activity not associated
with development, might improve model pre-
dictability.
Management of bald eagle foraging areas
should focus on maintaining and creating for-
ested shoreline as close as possible to the water.
In particular, shoreline trees >20 cm in dbh
should be protected, and dead trees should be
allowed to stand. The best way to provide a
sustained yield of suitable perch trees may be
to maintain intact patches of forest along the
shoreline. Our models can be used to identify
additional suitable shoreline areas in northern
Chesapeake Bay.
Given the increasing human population sur-
rounding Chesapeake Bay (Gray et al. 1988),
the future of the bald eagle population is seem-
ingly threatened with increased levels of human
disturbance and development. Models such as
ours should be used to map shoreline used by
eagles and areas with potential for use. Poten-
tially used areas should be preserved to accom-
modate the increase in eagle numbers as the
population grows. Restorable areas (principally
uplands with little vegetation) also should be
identified. Maps of suitable shoreline should be
used to plan habitat protection.
LITERATURE CITED
BUEHLER, D. A., S. K. CHANDLER, T. J. MERSMANN,
J. D. FRASER, AND J. K. D. SEEGAR. 1992. Non-
breeding bald eagle perch habitat on the north-
This content downloaded from 160.36.178.25 on Tue, 05 Jan 2016 16:18:33 UTC
All use subject to JSTOR Terms and Conditions

332 BALD
EAGLE SHORELINE PERCH USE * Chandler et al. J. Wildl. Manage. 59(2):1995
ern Chesapeake Bay, Maryland. Wilson Bull. 104:
540-545.
, T. J. MERSMANN,
J. D. FRASER,
AND
J. K. D.
SEEGAR. 1991. Effects of human activity on
bald eagle distribution on the northern Chesa-
peake Bay. J. Wildl. Manage. 55:282-290.
BYRD, M. A., G. D. THERRES, S. N. WIEMEYER, AND
M. PARKIN. 1990. Chesapeake Bay region bald
eagle recovery plan. U.S. Fish and Wildl. Serv.,
Newton Corner, Mass. 56pp.
CAIN, S. L., AND J. I. HODGES. 1989. A floating-fish
snare for capturing bald eagles. J. Raptor Res.
23:10-13.
CHESTER, D. N., D. F. STAUFFER,
T. J. SMITH,
D. R.
LUUKKONEN, AND J. D. FRASER. 1990. Bald
eagle habitat use in North Carolina. J. Wildl.
Manage. 54:223-234.
FIELDER, P. C., AND R. G. STARKEY. 1980. Win-
tering bald eagle use along the upper Columbia
River. Pages 177-194 in R. L. Knight, G. T. Al-
len, M. V. Stalmaster, and C. W. Servheen, eds.
Proc. Washington bald eagle symposium. The
Nat. Conservancy, Seattle, Wash.
FIENBURG,
S. E. 1980. The analysis of cross-classi-
fied categorical data. Second ed. MIT Press,
Cambridge, Mass. 198pp.
FRASER,
J. D., L. D. FRENZEL,
AND
J. E. MATHISEN.
1985. The impact of human activities on breed-
ing bald eagles in north-central Minnesota. J.
Wildl. Manage. 49:585-592.
GRAY, R. J., J. C. BREEDEN, J. B. EDWARDS, M. P.
ERKILETIAN,
J. P. BLASE
COOKE, O. J. LIGH-
THIZER, M. J. FORRESTER, JR., I. HAND, J. D.
HIMES, A. R. MCNEAL, C. S. SPOONER, AND W.
T. MURPHY,
JR. 1988. Population growth and
development in the Chesapeake Bay watershed
to the year 2020. U.S. Environ. Prot. Agency,
Chesapeake Bay Liaison Off., Annapolis, Md.
73pp.
GRIER,
J. W., J. B. ELDER, F. J. GRAMLICH,
N. F.
GREEN,
J. V. KUSSMAN,
J. E. MATHISEN,
AND
J.
P. MATTSSON. 1983. Northern states bald eagle
recovery plan. U.S. Fish and Wildl. Serv., Min-
neapolis, Minn. 121pp.
HARLOW,
W. M., E. S. HARRAR,
J.
W. HARDIN,
AND
F. M. WHITE. 1991. Textbook of dendrology.
Seventh ed. McGraw-Hill, New York, N.Y. 501pp.
KNIGHT, R. L., AND S. K. KNIGHT. 1984. Responses
of wintering bald eagles to boating activity. J.
Wildl. Manage. 48:999-1004.
MCGARIGAL, K., R. G. ANTHONY, AND F. B. ISAACS.
1991. Interactions of humans and bald eagles.
Wildl. Monogr. 115. 47pp.
MURPHY,
T. M., F. M. BAGLEY,
W. DUBuc, D. MA-
GER, S. A. NESBITT, W. B. ROBERTSON, AND B.
SANDERS. 1989. Southeastern states bald eagle
recovery plan. U.S. Fish and Wildl. Serv., Atlan-
ta, Ga. 94pp.
REVELATION
TECHNOLOGIES,
INC. 1991. Advanced
revelation version 2.12. Revelation Tech., Inc.,
Stanford, Conn. Unpaginated.
SMITH, T. J. 1988. The effect of human activity on
the distribution and abundance of the Jordan
Lake-Falls Lake bald eagles. M.S. Thesis, Vir-
ginia Polytech. Inst. State Univ., Blacksburg.
100pp.
STALMASTER, M. V. 1987. The bald eagle. Universe
Books, New York, N.Y. 227pp.
--, AND J. R. NEWMAN. 1978. Behavioral re-
sponses of wintering bald eagles to human activ-
ity. J. Wildl. Manage. 42:506-513.
- , AND - . 1979. Perch-site preferences
of wintering bald eagles in northwest Washing-
ton. J. Wildl. Manage. 43:221-224.
STEENHOF,
K. S., S. BERLINGER,
AND L. H. FRE-
DRICKSON.
1980. Habitat use by wintering bald
eagles in South Dakota. J. Wildl. Manage. 44:
798-805.
TITUS, K., J.
A. MOSHER, AND B. K. WILLIAMS. 1984.
Chance-corrected classification for use in dis-
criminant analysis: ecological applications. Am.
Midl. Nat. 111:1-7.
WALLIN, D. O., AND M. A. BYRD. 1984. Caledon
State Park bald eagle study. Unpubl. rep., Coll.
William and Mary in Virginia, Williamsburg.
53pp.
YOUNG,
L. S. 1983. Movements of bald eagles as-
sociated with autumn concentrations in Glacier
National Park. M.S. Thesis, Univ. Montana, Mis-
soula. 102pp.
Received 1 February
1994.
Accepted 27 November 1994.
Associate Editor:
Ryan.
This content downloaded from 160.36.178.25 on Tue, 05 Jan 2016 16:18:33 UTC
All use subject to JSTOR Terms and Conditions