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



351 



when ice moved substantially during the time that 

 elapsed between neighboring photographs. Seals in 

 overlapping zones were counted only once (i.e., only 

 counted in one of the overlapping images). In 2001, the 

 resolution in the low-altitude transects was superior to 

 that in the high-altitude transects; therefore we counted 

 all seals within the low-altitude imagery first and then 

 added counts of seals in the nonoverlapping portions of 

 the high-altitude imagery. Three replicate total counts 

 were calculated for each survey by counting the seals 

 in each image three separate times and tallying the 

 resulting counts for each replicate. All replicates were 

 counted by a primary analyst, and at least one week 

 elapsed between each replicate count to minimize bias 

 caused by the analyst remembering the location of seals 

 from previous replicate counts. The number of images 

 analyzed varied between surveys, depending on the 

 survey tracks flown and the distribution of ice. Sixty- 

 two total images were analyzed: 15 August 2001=21 

 images, 16 August 2001 = 30 images, and 15 August 

 2002=11 images. A subsample of five images from Johns 

 Hopkins Inlet was also counted by a secondary analyst 

 to provide independent verification of counts. The mean 

 count (of the three counts by the primary analyst) for 

 each image was calculated, and the mean counts for 

 all images from each survey were summed to estimate 

 the total survey count. The variance of the total count 

 estimate was estimated as the sum of the variances for 

 each mean count included in the total estimate. 



Comparison of detection rates at different altitudes 



To compare the detection rates in low- and high-altitude 

 imagery from 2001, we counted all seals in the over- 

 lapping zones of low- and high-altitude images. Next, 

 we visually compared the location of each seal in the 

 overlapping portions of low- and high-altitude images 

 and classified seals as either 1) counted in both images 

 or 2) counted in only one of the images. Seals counted 

 in only one of the images were further categorized as 

 follows: 1) light-colored seal not detected in the other 

 image, 2) seal in a group not resolved as an individual 

 in the other image, 3) seal definitely not present in the 



other image (e.g., seal went into the water or hauled out 

 between transects), or 4) shadow or dirty ice classified as 

 a seal in other image. These comparisons were conducted 

 to help us understand the relative accuracy of count- 

 ing seals from images taken at different altitudes. The 

 less accurate high-altitude counts were not used when 

 estimating mean counts for each survey; mean counts 

 were estimated by triplicate counts with priority given 

 to low-altitude counts as described above. 



Results 



Comparison of total counts 



In 2001, counts made from shore were consistently 

 higher than counts made with the use of aerial pho- 

 tography (Table 1). In contrast, both counting methods 

 produced similar results in 2002. The standard errors 

 and coefficients of variation (CV) presented in Table 1 

 reflect variance between counts by shore-based observers 

 or between independent counts of aerial photographs. 

 Although the CVs for counts of individual images were 

 generally larger than the CVs for total estimates, 91% of 

 the CVs for individual images were less than 0.1. Of the 

 five images counted by a secondary analyst, all counts 

 were within 87c of the mean of the three replicate counts 

 conducted by the primary analyst. Imprecision or inac- 

 curacies in counts caused by the distance of seals from 

 the observation site or the altitude of the aerial survey 

 were not easily quantified, although altitude-related 

 errors were evaluated separately by comparing counts of 

 seals in overlapping low- and high-altitude images. 



Spatial distribution of seals in Johns Hopkins Inlet 



The distribution of seals in Johns Hopkins Inlet was 

 different during each of the surveys and appeared to be 

 associated with the pattern of ice in the inlet. Gener- 

 ally, seals were found in aggregations, although solitary 

 seals were frequently observed outside of the main seal 

 concentrations (Fig. 3). On 15 August 2001, seals were 

 distributed in groups of 200-500 animals, ranging 



