190 
Fishery Bulletin 1 14(2) 
zone for monitoring during each 10-min period. At the 
start of each 10-min period, the observer chose a gen- 
eral area within the selected zone as the initial search 
area for scans. We tried to vary the ice type (i.e., either 
dense, with icebergs packed together, or scattered, with 
much open space among icebergs) and the position of 
the initial search area in subsequent samples. 
Once an initial search area was selected, the ob- 
server used binoculars or a spotting scope and the 
first seals to come into view became the focal seals for 
that period. If possible, we added additional seals in 
the immediate vicinity to try to bring the total number 
of focal seals to between 6 and 10. We recorded the 
size of the seal group, number of mother-pup pairs in 
each group, ice type, the position of the seals in rela- 
tion to an overall patch of icebergs (e.g., interior versus 
near an edge next to a section with open water or less 
ice), and weather (i.e., sky condition, precipitation, and 
wind speed). During each 10-min observation period, 
we recorded the number of seals that entered the water 
and the number of vessels in the inlet. In 2001, we also 
recorded the vessel type of the nearest vessel and the 
closest approach (in the categories 0-50 m, 51-100 m, 
101-300 m, >300 m) of that vessel to the focal seals. 
Observations were terminated early if any of the focal 
seals drifted out of view; these rare, incomplete obser- 
vations were not used. 
We used a generalized linear mixed model (GLMM; 
binomial error, logit link) to model how the probability 
of a seal entering the water was affected by external 
factors. The response variable was the number of seals 
that entered the water in the 10-min observation peri- 
od in relation to the number of seals under observation. 
The predictors were DOY, TOD, seal position on the 
ice, number of mother-pup pairs, ice type, sky condi- 
tion, precipitation, wind speed, number of vessels, ves- 
sel type, and distance to vessels. We included quadratic 
terms for DOY and TOD to allow for nonmonotonic as- 
sociations. For several variables (e.g., sky condition, 
precipitation, wind speed, vessel distance), categories 
with few observations were combined with adjacent 
categories. The predictors mother-pup pairs and vessels 
were, in separate models, fitted as either continuous 
predictors (counts) or binary predictors (presence or ab- 
sence). Vessel type was fitted as a single, multicategory 
predictor to determine whether responses differed by 
vessel type; then we fitted models for the presence of 
each vessel type separately to produce individually es- 
timated odds ratios for each vessel type. 
Initially, models included DOY, TOD, their quadratic 
terms, seal position on the ice, and all weather vari- 
ables. Wald chi-square statistics (P>~0.05) were used 
to eliminate unimportant variables one at a time (Hos- 
mer and Lemeshow, 2000). After this model fitting (i.e., 
after we determined useful environmental predictors), 
mother-pup pairs and vessel predictors were included 
and evaluated in the same way. 
Vessel approaches To estimate the effect of vessel type 
and distance more directly on seal behavior (i.e., the 
probability that a seal will enter the water), in 2001 
we observed a subset of the vessels entering Tracy Arm 
as they approached seals hauled out on icebergs. When 
seals are hauled out, they are in a temporary state that 
ends when the seal enters the water. As such, external 
factors, including the presence of boats, can only affect 
the timing of when a hauled-out seal enters the water, 
not the fact that the entry will occur. We used entry 
into the water as our response because it is an eas- 
ily observed and unambiguous behavior with energetic 
and predator-exposure consequences, and it has been 
used as a measured response in most comparable stud- 
ies (Calambokidis et al. 1 ; Mathews, 1995; Jansen et al., 
2010; Young et al., 2014). 
Once a vessel was selected for observations of seal- 
vessel interaction, the observer selected one or more 
focal icebergs in the vessel’s path and recorded the 
number of nonpups and pups on each iceberg. We esti- 
mated the distance between the vessel and seals and 
classified it into 1 of 4 distance classes: 0-50 m, 51-100 
m, 101-300 m, and >300 m. When possible, we selected 
icebergs that were >300 m from the vessel at the start 
of an observation, but not all distance classes were ob- 
served for every vessel-iceberg combination. For each 
distance class transited by a vessel, we recorded the 
number of pups and nonpups that entered the water 
(which could be zero). We continued to monitor icebergs 
until shortly after the vessel made its closest approach 
to the seals, at which point observations were termi- 
nated and a new iceberg was selected for focus. We at- 
tempted to monitor a vessel during its entire inbound 
and outbound track, and we recorded both seal behav- 
ior and vessel activity (i.e., vessel moving or not mov- 
ing) during these observations. Most observations were 
made within 1 km of the observation point. 
We evaluated our visual distance estimates by 
comparing them with measurements made with laser 
rangefinder binoculars (Leica Vector IV 8 , Vectronix, 
Inc., Bedford, NH), which allowed us to measure the 
distances to the vessel and to the seal(s) and the angle 
between them, automatically calculating the distance 
between the vessel and the seal(s); the rangefinder 
was calibrated according to the manufacturer’s recom- 
mendations. For this effort, 2 observers simultaneously 
recorded distances between selected icebergs or an ice- 
berg and a vessel; 1 observer estimated the distance 
visually, using the seal-vessel distance classes, and the 
other observer measured the distance with the laser 
rangefinder. The distance measurements made with the 
rangefinders were then converted to distance classes, 
and the visual estimates of distance were scored into 3 
bins: classified correctly, overestimated by 1 or 2 class- 
es, or underestimated by 1 or 2 classes. 
Although we did not systematically monitor such 
behavior, we occasionally observed seals entering the 
water after the vessel had passed its point of closest 
Mention of trades names or commercial companies is for 
identification purposes only and does not imply endorsement 
by the National Marine Fisheries Service, NOAA. 
