Ball et al. • NOCTURNAL PROVISIONING OF THRUSHES 
509 
northwest of Manning, Alberta (57°18'N, 
118"23' W) in 2006 and 2007. The Fort Simpson 
study site was >500 km north of Chinchaga and 
had —19.7 hrs of daylight at the summer solstice 
compared to 18 hrs of daylight at Chinchaga. 
Field Procedures. —We visited each study plot 
every 3 days between late May and mid-July to 
search for and to monitor nests. We randomly 
selected 37 Swainson’s Thrush nests for video 
monitoring and placed an infrared digital video 
camera at each nest to continuously and unobtru¬ 
sively record parental behaviors. We observed 
each brood at —4 and 8 days of age to account for 
age-related changes in nestling energy demand 
(Weathers 1996). Age estimates for each brood 
were based on video footage and hatch date 
estimates from repeat nest visits. We viewed nine 
l-hr time periods during each age. Six diurnal 
periods had start times evenly distributed between 
1 hr post-sunrise and 2 hrs pre-sunset. Three 
nocturnal periods had start times at sunset, 2 hrs 
post-sunset, which was approximately the middle 
of the night, and 1 hr prior to sunrise (hereafter 
post-dusk, night, and pre-dawn, respectively). We 
quantified provisioning rate (number of food 
deliveries to the nest/hr), nestling age as a 
continuous variable, and the average number of 
nestlings present for each lime period. We were 
unable to consistently identify the size of prey 
items and no attempt was made to estimate rales 
of biomass or energy delivery. Nest fate and 
number of young fledged were ascertained from 
video footage, A nest was considered successful if 
-I nestling fledged, which we defined as 
departing the nest under its own power. Age of 
fledge refers to age at which the first fledgling left 
the nest. 
We quantified three covariates that we expected 
roight influence nocturnal provisioning. Canopy 
tree density/m 2 reflects habitat complexity and 
potential light availability to the forest understory. 
We counted the number of trees s3 m in height 
within an 11.3-m radius around each nest (Martin 
et al. 1997). We expected large openings in the 
forest canopy to similarly influence nocturnal 
visibility and navigability. We used ArcGIS 
(ESRI 2009) to calculate the distance between 
each nest and the nearest canopy edge created by 
a pipeline, seismic line, road, or river. Finally, we 
expected nestling energy demand and the need for 
provisioning to increase during colder tempera¬ 
tures. We placed a weather station (ONSET 2006) 
at the center of each study area under a forest 
canopy similar to the study plots and measured 
temperature at 5-min intervals. Temperature data 
were averaged/hr for each time period that 
provisioning was observed. 
Statistical Analyses .—All covariates were test¬ 
ed for normality using normal probability plots 
and a combined test of skewness and kurtosis 
(D’Agostino et al. 1990, Royston 1991, Zuur et al. 
2009). Nestling number was squared to achieve 
normality (x ‘2 = 9.23, P = 0.89). We applied 
square root and log transformations to edge 
distance and tree density, respectively, to improve 
skewness and kurtosis. Temperature, day length, 
and nestling age were not transformed. 
We used mixed-effect Poisson regression 
models in an information-theoretic framework 
using A1C £ . to analyze variation in hourly 
provisioning rate (Burnham and Anderson 2002). 
Nest identity was included as a random effect in 
al) models to account for the repeated measure of 
foraging rate at different ages. We calculated 
AIC lt . and evidence ratios to compare the relative 
support among models (Burnham and Anderson 
2002, Anderson 2008). Our base hourly provi¬ 
sioning rate model represented brood demand and 
assumed that hourly provisioning rate was a 
function of nestling age and nestling number. 
We compared support for our base model to 
models that considered the effect of time period, 
study area, and an interaction between time period 
and study area. 
We compared diurnal provisioning rates be¬ 
tween pairs that did and did not provision at night 
to examine if night provisioners were compensat¬ 
ing for an energy shortfall or were delivering 
bonus energy. We compared support for the top 
hourly provisioning rate model to models that 
included a binary variable identifying pairs as 
night provisioners or non-night provisioners, 
which we added either as a main effect or as an 
interaction with time period. These analyses were 
performed on a subset of data that excluded the 
night period. Support for the interaction would 
indicate night provisioners fed their young at a 
different rate during the day compared to parents 
that did not provision at night. 
We compared support among six models to 
explain variation in provisioning rate during the 
night period. Five models each considered the 
individual effects of brood demand (base model), 
mean hourly temperature, day length, canopy tree 
density, or distance to forest edge. The sixth 
model simultaneously considered the above five 
