Evert et al. • DISTRIBUTION OF MIGRANT LANDBIRDS ALONG LAKE HURON 539 
compare microclimatic differences between 
shoreline and inland areas, and potential food 
resources dependent on foliage. We collected 
phenology data following each 5-min bird survey 
on 10 paper birch and 10 quaking aspen trees 
nearest the center of the point-count station. 
Spring phenology was characterized as: (I) leaves 
in bud: (2) leaves emerging from bud (bud scales 
shed). (3) leaves unfurled but not fully expanded, 
and (4) leaves fully expanded. We categorized fall 
phenology as: (1) little or no loss of green leaves, 
(21 retention of >50% of leaves, some of which 
may have lost chlorophyll, (3) retention of <50% 
of leaves, some of which may have lost 
chlorophyll, and (4) loss of all leaves. 
Midge abundance was visually estimated as the 
number of insects on vegetation and in the air 
within a 5-m radius circle from the center of the 
point-count circle. We quantified midge abun¬ 
dance during a 10-sec period following each bird 
count into one of four categories: (I) no midges. 
(2) 1-500 midges. (3) 501-1,000 midges, and (4) 
1.000 midges. We did not collect data on other 
arthropod groups because we noted few other 
arthropods. Midge data were collected during 
spring and fall 1994 after it became evident birds 
were foraging on them in 1993, the first year of 
the study. 
Statistical Analyses .—Distributions of relevant 
variables were examined for departures from 
normality and non-parametric statistics were used 
when transformations did not bring data into 
compliance with parametric testing assumptions 
(Zar 1996), Bird and midge count data were 
analyzed using SPSS 16.0 (SPSS 2008). We used 
general linear models (GLM) for natural log 
'ransformed data [Y = Ln(X + 1)J, or GLM for 
ranked data followed by Tukcy post hoc tests, 
Kniskal-Wallis Analysis of Variance, and Mann- 
Whitney tests (Conover 1999). Wc used Kendall 
partial rank-order correlations, a nonparamelric 
method that eliminates the effect of a third 
variable on the relationships between the variables 
°f interest, when necessary (Siegel and Castellan 
1988). 
Bird counts were modeled as a function of 
Perpendicular distance from shoreline. We includ¬ 
ed transect as a factor in the analyses to control 
for effects of both longitude and individual 
transect, two potentially confounding variables. 
We included three temporal variables in each 
analysis: year, season, and date. Reported values 
are means ± SE. 
(A) 
FIG. 2. Mean (± SE) number of birds detected per 
point as a function of distance from the northern Lake 
Huron shoreline for spring (A) and fall (B) migration, 
respectively, 1993-1994 pooled. 
RESULTS 
Birds .—We recorded 2,577 observations of 
birds during the 2-year study (Appendix). Birds 
were most abundant near the shoreline for both 
long- and short-distance migrants (Fig. 2). 
Spring.—Abundance of long-distance migrants 
was affected by year (F|, 8 08 = 21.01 . P < 0.001), 
distance from shore (F 4 . 8 O 8 = 1 1-06. ^ < 9.001), 
transect (F 88(m = 4.14, P < 0.001), and date 
(Fi a .808 = 72.40, P < 0.001). Sites within 0.4 km 
of the shoreline had more birds than sites at 
distances greater than 0.4 km from the water 
(Fig. 2A). There was an immediate shoreline 
effect with more birds counted at the shoreline 
relative to all other distances except 0.4 km. There 
was no significant difference between number of 
birds observed at the immediate shoreline and 
birds counted at a distance of 0.4 km (Tukey P = 
