Mathews et al.: Effects of vessel disturbance on Phoca vitulina on glacial ice in Tracy Arm, Alaska 
191 
approach to a seal or seal group because of wake ef- 
fects or for other reasons (also reported by Calamboki- 
dis et al. 1 ). If the seal entered the water when the ves- 
sel was still in the same distance class as the closest 
approach, we used the observation in the analyses, as- 
suming that most such disturbances would have been 
observed. However, we did not use observations of dis- 
turbances when vessels had moved beyond the distance 
class of their closest approach because we were far less 
likely to observe them. Consequently, our estimates of 
the probability of disturbance are biased low but are 
comparable to those of other studies where seals were 
not tracked after vessels passed them (Jansen et al., 
2010; Young et al., 2014) and comparable to counts of 
“visual reactions” (versus “wake reactions”) reported by 
Calambokidis et al. 1 
We used a GLMM (binomial error, logit link) to esti- 
mate the effect of predictor variables on the probability 
of a seal entering the water. We used the following pre- 
dictors: distance between the seal and vessel, seal age 
class (pup, nonpup), DOY, TOD, group size (the number 
of seals on an iceberg when the vessel entered each dis- 
tance zone), percent ice cover in the area of the moni- 
tored iceberg, if >1 pup was on an iceberg when it was 
first observed, vessel activity (i.e. , moving, not moving), 
sky condition, temperature, and precipitation. Encoun- 
ter (i.e., individual vessel observed) and iceberg within 
an encounter were included in the model as random ef- 
fects because observations within these units were not 
independent. We included quadratic terms for DOY and 
TOD to allow for nonlinear (on the logit scale) respons- 
es between the predictors and water-entry probability. 
We also included the interaction terms distance 
class*vessel type and distance class*age to allow the 
effect of distance class to vary by vessel type or by seal 
age category. The effect of age and distance class*age 
were investigated by using a data set that separated 
observations of pups and nonpups; a data set that in- 
cluded both pups and nonpups was used to investigate 
all other predictors. For 2 distance*vessel categories 
(skiff: >300 m; kayak: >300 m), there were no observa- 
tions of seals entering the water — a situation that pre- 
vented successful model fitting and estimation of pa- 
rameters (Hosmer and Lemeshow, 2000). To overcome 
this problem and facilitate model fitting, we added 0.1 
(i.e., added one-tenth of a water entry) to one observa- 
tion in each category. Because of this adjustment, the 
estimates for these 2 categories are slightly positively 
biased, although they remain essentially zero. 
Results 
Seal counts and timing of pupping 
In 2001, we counted seals in Tracy Arm 193 times on 34 
days (range: 3-10 counts per day); 100 of these counts 
were paired (50 pairs), with counts differing by >10% 
in 8 of the 50 pairs. All counts were included in the 
analyses. Total counts (i.e., pups+nonpups) peaked dur- 
ing 24-26 June, with a high count of 1351 seals (Fig. 
2A). Separate counts of both pups and nonpups peaked 
during this period, at -400 pups and -1000 nonpups 
(Fig. 2B). The average numbers of pups seemingly be- 
gan to level off by about 25 June, with the maximum 
pup count of 408 observed on June 24; however, it is 
unclear whether the average nonpup numbers were 
still increasing or were at a peak on 30 June, the last 
day that we made observations in 2001 (Fig. 2B). We 
observed the first evidence of birth (blood on iceberg) 
on 30 May, the third day of our study, and we recorded 
the last on 25 June. Peak daily numbers of icebergs 
with evidence of recent births (4-8) were observed dur- 
ing 7-13 June. 
For all 3 groupings of seals (nonpups, pups, and all 
seals), 6 variables (count quality, DOY, TOD’TOD, per- 
cent ice cover, sky condition, and temperature) were re- 
tained in the GLMs (Table 1). Mean counts were low- 
est for high count qualities and highest for intermedi- 
ate count qualities (Table 1). The number of nonpups 
peaked at around 1300, whereas the number of pups 
showed no distinct diurnal pattern (Fig. 3). Seal counts, 
for both pups and nonpups, were positively related to 
ice cover (Table 1). More seals were counted under 
clear skies, more nonpups were counted when there 
was no precipitation, and more pups were counted 
when wind speed was intermediate among the condi- 
tions we observed (i.e., Beaufort 2) compared with the 
counts when there was more (i.e., Beaufort 3) or less 
(i.e., Beaufort 1) wind (Table 1). Counts of nonpups and 
pups increased with increasing temperatures (Table 1). 
Vessel traffic 
The mean numbers of vessels observed daily in Tracy 
Arm varied from 10.2 in 2001 to 2.0 in 2006 (Fig. 4). 
The maximum number of vessels recorded in a day was 
33 on 26 June 2001. Also in 2001, the mean number of 
vessels per day on weekdays (10.25, n=24) was similar 
to that observed on weekend days (10.20, n=10); conse- 
quently, this factor (i.e., weekday versus weekend day) 
was not used in other analyses. 
Tour boats were the most common type of vessel ob- 
served in most years, followed by power boats (Fig. 4), 
and these 2 types of boats accounted for 57-100% of 
the vessels observed in a day. The number of sampling 
days was small during the years 2004-2006, making 
the ranking of vessel types potentially imprecise for 
those years. The majority of inflatables, skiffs, and kay- 
aks were launched from larger vessels. 
In 2001, most vessels (87%) entered section A, the 
part of our study area closest to the glacier (Fig. 1, in- 
set map). Individual vessels were in the study area an 
average of 1.2 h both in 2001 (n= 269; 95% confidence 
interval [Cl]: 1.10-1.30) and during 2002-2006 (n=193; 
95% CI:1.12— 1.30). Some estimates of mean lengths of 
stay were underestimated when vessels arrived or de- 
parted outside of our observation periods. The number 
of vessels per day in Tracy Arm declined over the pe- 
riod 2001-2006, both in total and for each vessel type 
