194 
Fishery Bulletin 1 14(2) 
Time of day 
Figure 3 
Counts of pup (p) and nonpup (np) harbor seals (Phoca vitulina), including 
seals that were hauled out and those in the water, in Tracy Arm, Alaska, from 
27 May-30 June 2001 as a function of time of day (TOD). Lines are the mod- 
eled functions of the mean counts in relation to TOD and are adjusted for 
other predictors in the models. 
sels approached. Vessels were observed for an average 
of 24 min (standard deviation=29 min). Individual ice- 
bergs had 1-15 nonpup seals and 1-5 pups; 725 ice- 
bergs (-60% of those observed) had at least 1 pup. Dis- 
turbances caused by monitored vessels occurred on all 
of the 32 days we observed vessel approaches. 
Overall, 74% of visual estimates of distance were 
correctly classified by observers. Observers tended to 
estimate shorter distances more accurately than lon- 
ger distances (93% of distances <50 m were classified 
correctly, but only 64% of distances >300 m were cor- 
rectly categorized). Observers tended to underestimate 
rather than overestimate distances, and this bias in- 
creased with distance. Overall proportions of misclassi- 
fication were similar for both observers, although vari- 
ability between observers also increased slightly with 
distance. 
Day of year and TOD were not related to the proba- 
bility of a seal’s entry into water (water entry [P>0.35]), 
nor were seal age category (including age*distance 
class), temperature, or precipitation (Fig. 6). Predictors 
associated with the probability of water-entry included 
seal group size, the presence of a pup at the start of 
the observation, percent ice cover, sky condition, vessel 
activity, and vessel distance, the latter 2 of which var- 
ied by vessel type (Figs. 6 and 7). Seals were more like- 
ly to leave an iceberg as ice cover decreased or under 
clear versus cloudy skies (Fig. 6). Seals were less likely 
to leave an iceberg with increasing group size but were 
much more likely (1.3 times [95% Cl : 1.0-1. 8]) to enter 
the water if there was at least 1 pup present in the fo- 
cal group of seals on an ice berg when the observation 
began (Fig. 6, pup present). However, pups were rarely 
the first seal in a group to enter the water. In 56 of 
the 278 (20%) instances when a seal entered the water 
when a pup was present, we recorded whether a pup 
or a nonpup entered the water first (in 222 instances, 
both entered at the same time or the order was not 
clear from the recorded data). For 43 of the 56 (77%) 
instances, a nonpup entered the water first. 
The odds of a seal entering the water when a ves- 
sel was stopped were about 4 times as high as when a 
vessel was moving (Fig. 6, vessel activity). The prob- 
ability that a seal entered the water increased dra- 
matically as the distance between a seal and a vessel 
decreased, but the effect differed depending on vessel 
type (Table 2, Fig. 7). The probability of a seal respond- 
ing to a vessel was higher for cruise ships and kayaks 
than for other vessel types (Fig. 7). The probability of 
a seal entering the water when vessels were within 50 
m was >0.47 for every vessel type and almost 1 for 
cruise ships and kayaks. For vessel types other than 
cruise ships and kayaks, the probability of water entry 
by seals decreased to low levels for distances >100 m. 
Probabilities of water entry were uniformly low for all 
vessel types at distances >300 m (Fig. 7). 
Discussion 
Seal counts 
In Tracy Arm in 2001, there were more seals (both 
nonpups and pups) hauled out with greater ice cover, 
