352 



Fishery Bulletin 105(3) 



from 1 to 6 km from the glacier terminus (Fig. 3A). On 

 16 August 2001, a large group of 800-900 seals was 

 aggregated in a band stretching from 0.5 km from the 

 glacier terminus to the shore-based observation site, and 

 another group of 300-400 seals was found 3-5 km from 

 the glacier (Fig. 3B). On 15 August 2002, all of the seals 

 were within 3 km of the glacier face between the glacier 

 and the shore-based observation site (Fig. 3C). Of these, 

 a group of 350-450 seals was observed on the southwest 

 side of the inlet, and the remaining 1100-1200 seals 

 formed a dense concentration on the east side of the 

 inlet. No seals were located in the blind areas (from the 

 view point of the shore-based observers) on any of the 

 three survey days (Figs. 2A and 3). 



dirty ice, 0.5% were light-colored seals that were not 

 detected, and 0.3% were so close to other seals that 

 they could not be identified from their neighbors). In 

 the 16 August 2001 survey, 34.3% of the seals counted 

 in the low-altitude images were misclassified in the 

 high-altitude images, including 12.5%' that were shad- 

 ows or dirty ice misidentified as seals and 21.8% that 

 were not detected in the high-altitude imagery (21.5% 

 dark seals and 0.3% light-colored seals). The net effect 

 of the misclassifications was that counts from higher- 

 altitude images were underestimates of the number of 

 seals compared to counts from lower altitudes (i.e., the 

 proportion of seals missed exceeded the proportion of 

 false identifications). 



Comparison of detection rates at different altitudes 



An examination of seals in overlapping zones of low- and 

 high-altitude images revealed a difference in the rates 

 of seal detection between the two altitudes. Seals were 

 more easily detected and confirmed to be seals in the 

 low-altitude imagery; therefore, seals identified in low- 

 altitude imagery were considered to be "true" observa- 

 tions for comparison to seals counted in high-altitude 

 imagery. For the 15 August 2001 survey, 32.7%' of the 

 seals counted in the low-altitude images (i.e., "true" 

 seals) were misclassified in high-altitude images: 8.6% 

 were counted as seals when no seals were present (i.e., 

 shadows or dirty ice misidentified as seals) and 24.1% 

 were not counted as seals when seals were present 

 (i.e., 23.3% were dark seals dismissed as shadows or 



A 15 Aug 2001 



16 Aug 2001 



IN 



\ 



C 1 5 Aug 2002 



Discussion 



Comparison of total counts 



Both shore-based and aerial counts indicated that 

 more than 1500 seals haul out on glacial ice in Johns 

 Hopkins Inlet in mid-August, making the inlet one of 

 the most important haul-out sites in Glacier Bay (as 

 suggested by Mathews [1995]). The total number of 

 seals that use the inlet might be substantially larger 

 because some unknown proportion of seals was in the 

 water (i.e., not hauled out on ice) during the surveys. 

 In 2001, counts made from shore were consistently 

 higher than counts made with aerial photography 

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

 similar results in 2002. Several sources 

 of error for each method likely contrib- 

 uted to these inconsistencies in results 

 between the two methods. 



glacier 



glacier 



glacier 



5 km 



Figure 3 



Relative distribution of patches of harbor seals (Phoca vitulina) in 

 Johns Hopkins Inlet, Glacier Bay, determined by aerial photography, 

 on (A) 15 August 2001, (B) 16 August 2001, and (C) 15 August 2002. 

 Dots indicate groups of seals and are meant to illustrate generalized 

 locations of seals, rather than a precise indication of seal abundance. 

 The shore-based observation site is indicated with a star. 



Sources of error for each survey method 



Both counting methods were susceptible to 

 common errors of either double-counting 

 or missing seals. These errors were most 

 likely to occur within overlapping zones 

 between neighboring photographic images, 

 between parallel passes with binoculars, 

 or between shore-based counts of subar- 

 eas. If overlapping zones were not accu- 

 rately delineated, individual seals within 

 the overlapping zone could be counted 

 twice, or missed entirely. The permanent 

 record provided by photography provided 

 the best opportunity to minimize such 

 errors by allowing for careful delineation 

 of overlapping zones based on the relative 

 positions of identifiable pieces of ice on 

 adjacent images. The shore-based method 

 did not allow re-identification of individual 

 pieces of ice; therefore shore-based observ- 

 ers attempted to eliminate overlapping by 

 adjusting binoculars carefully. Seals could 

 be missed, however, if the binoculars were 



