Bengtson et al.: Comparison of survey methods for estimating abundance of Phoca vitulina 



353 



lowered more than one field of view. Similarly, seals 

 could be counted twice, if the binoculars were lowered 

 less than one field of view before the second survey pass. 



Counting errors could be caused by movement of ice 

 on which seals were hauled out. Shore-based observers 

 could not track the movement of ice between paral- 

 lel passes; therefore some counting errors caused by 

 ice movements were unavoidable. Depending on ice 

 drift patterns, seals that were already counted could 

 drift into an uncounted zone and be double counted, 

 or uncounted seals could drift into a counted zone 

 and be missed. On both days surveyed in 2001, we 

 observed considerable movement of ice between aerial 

 transects, particularly along the eastern side of the 

 inlet, farthest away from the shore-based observation 

 site. Thus, it is likely that seals would have drifted 

 between the shore-based observers' counting areas, 

 resulting in either missed or double counts. In con- 

 trast, the ice was less mobile during the 2002 survey 

 day. Ice did not drift much between adjacent photo- 

 graphic images along a transect because only 5-10 

 seconds elapsed between each image. However, ice 

 sometimes drifted substantially between images along 

 neighboring transects, which were typically sepa- 

 rated by 10-15 minutes. Although such ice movements 

 sometimes made identification of individual seals be- 

 tween neighboring images more difficult, spatial clues 

 from recognizable pieces of ice aided identification 

 and made us confident that seals on moving ice were 

 properly counted. 



The distribution of seals could also influence count- 

 ing errors for both survey methods. In 2001, ice was 

 distributed up to 4.5-6 km from the glacier terminus 

 and seal distribution was clumped (Fig. 3, A and B). 

 In contrast, during the 2002 survey day the ice was 

 more densely packed near the glacier terminus and 

 seals were more evenly distributed (Fig. 3C). The 

 distance between the shore observers and seals was 

 also expected to affect shore counts, and to produce 

 greater error as distance increased. Our results, how- 

 ever, did not exhibit a clear pattern in the percent 

 difference between counting methods and distance be- 

 tween shore observers and seals. In 2001, seals were 



within 4.25 km of observers on 15 August and within 

 2.4 km on 16 August; during the 2002 survey seals 

 were located within 2.75 km (Table 2). Overall, the 

 counts recorded by both methods were most similar 

 when ice movement and seal clumping were minimal 

 (Table 2). 



Counting errors could also be caused by misidenti- 

 fying seals as shadows or dirty ice. Occasionally, ice 

 ridges cast shadows that looked remarkably like seals. 

 Some glacial ice contained veins of dirt that also had 

 the similar shape and color of seals. When compar- 

 ing seals identified in overlapping imagery from the 

 two aerial survey altitudes, we found that 22-24% 

 of seals counted in the low-altitude (high-resolution) 

 imagery either were not detected or were dismissed 

 as shadows or dirty ice in counts of the high-alti- 

 tude imagery. However, 9-13% of seals counted in the 

 high-altitude imagery were actually shadows or dirty 

 ice, according to the low-altitude imagery. These two 

 types of errors tended to offset each other, although 

 high-altitude counts still exhibited a general bias 

 toward 10-15% lower seal counts. As a result, the 

 aerial counts (especially from high-altitude imagery) 

 likely produce underestimates of the actual number of 

 seals hauled out on ice. The aerial counts from 2001 

 were based primarily on low-altitude counts, and, thus 

 probably represent less biased estimates than the 

 counts for the 2002 survey, which was based solely 

 on high-altitude imagery. No correction was applied 

 to the high-altitude counts because the likelihood of 

 misidentifying or missing seals varied according to 

 conditions specific to each image (e.g., dirty ice and 

 shadows); therefore applying a correction based on 

 images from one part of the fjord probably was not 

 applicable to images from other parts of the fjord, let 

 alone images from an entirely different day. Further, 

 no correction was available for the proportion of seals 

 misidentified or missed in the low-altitude imagery 

 and for an unknown number of seals that were in the 

 water during the surveys. 



Shore-based observers had the advantage of a three- 

 dimensional "live" view of seals and were able to dis- 

 tinguish between actively moving seals and shadows 



