350 



Fishery Bulletin 105(3) 



binoculars were pivoted horizontally in one direction. 

 To facilitate systematic counting in the study area, 

 observers visually divided the field of view into three 

 or four subareas using markers such as landmarks or 

 natural breaks in the ice as viewed from the observa- 

 tion site. When a marker came into the field of view, 

 the binoculars were lowered exactly one field of view, 

 locked again, and a pass in the opposite direction was 

 made (Mathews, 1995). Each of the subareas typically 

 required only two nonoverlapping, parallel passes across 

 ice habitat to completely cover the width of a subarea. 

 Counts from all subareas were summed for each ob- 

 server to estimate total counts, and the two observers' 

 total counts were then averaged to estimate the total 

 count for each survey. The variance for each survey's 

 total count was estimated as the variance among the 

 two observers' total counts. 



Aerial surveys 



Aerial surveys were conducted from a twin engine air- 

 craft on 15 and 16 August 2001 and 15 August 2002 by 



glacier 



glacier 



glacier 



Figure 2 



Survey coverage of Johns Hopkins Inlet, Glacier Bay, achieved by each 

 survey type in this study. (A) Field of view (shaded area) observed 

 from shore-based observation site (indicated with a star). Note that 

 the northwestern edge of the inlet and a small area near the glacier 

 are obscured from view ("blind spots") by geographical features. The 

 outer range of visibility varied with conditions and is denoted by a 

 dashed, curved line. (B) Photographic coverage of aerial transects 

 flown on 15 and 16 August 2001. The straight lines in the inlet reflect 

 the approximate boundaries between photographs. For simplicity, only 

 the nonoverlapping areas (i.e., no endlap or sidelap) of each image are 

 shown. The central transect was flown at a lower altitude (915 m) 

 than the two adjacent transects (1465 m). (C) Photographic coverage 

 of aerial transects (1465 m altitude) flown on 15 August 2002; the 

 overlap between neighboring photographs was ignored. 



a commercial photographic surveying company (Aeromap 

 U.S., Anchorage, AK). In 2001, each aerial survey over 

 Johns Hopkins Inlet was completed in three transects: 

 two transects at 1465 m altitude covering the entire 

 inlet, and an additional transect at 915 m covering the 

 central portion of the inlet a second time from Johns 

 Hopkins Glacier to the point north of Jaw Point (Fig. 

 2B). In 2002, the survey aircraft flew two transects at 

 1465 m covering Johns Hopkins Inlet from the glacier 

 to Jaw Point; a lower-altitude, central transect was 

 not flown (Fig. 2C). Shore observers observed no reac- 

 tion by the seals (e.g., entering the water) to the aerial 

 surveys. 



During each survey, large-format (23x23 cm) photo- 

 graphic images were taken automatically at a predeter- 

 mined rate on Agfa Aviphot Color XlOO PEl negative 

 film (Agfa Corp., Ridgefield Park, NJ), with a belly- 

 mounted Zeiss RMK TOP 15 camera (Carl Zeiss, Inc., 

 Thornwood, NY) with forward-motion compensation 

 (15 August 2001) or with a Zeiss RMK A 15/23 camera 

 (Carl Zeiss, Inc., Thornwood, NY) (16 August 2001 and 

 15 August 2002). We did not notice any substantial im- 

 provement with the use of forward-motion 

 compensation. The resulting photograph- 

 ic frame widths (i.e., the on-the-ground 

 width of the area photographed in each 

 photo frame) were 2200 m for high-al- 

 titude (1465 m) images and 1400 m for 

 low-altitude (915 m) images. The images 

 had approximately 10% endlap (i.e., over- 

 lapping duplication of the same area in 

 separate, successive images) within tran- 

 sects, 20% sidelap between high-altitude 

 transects, and 75% sidelap between the 

 central transect and the neighboring 

 high-altitude transects in 2001. Large- 

 format negatives were scanned at 1600 

 dpi (the maximum resolution available to 

 us) with a digital scanner and converted 

 to positive-color digital images. The pixel 

 resolutions of the resulting digital images 

 were 0.10 m and 0.15 m for low- and high- 

 altitude transects, respectively. At that 

 resolution, we found that seals could be 

 identified from the scanned imagery, and 

 we were satisfied with this resolution for 

 the purposes of counting seals. 



Seals were counted from the digital im- 

 ages by using the "geospatial light table" 

 feature of ERDAS Imagine 8.6 software 

 (Leica GeoSystems Inc., Atlanta, GA). 

 No distinction was made between adults, 

 pups, or juvenile seals. A virtual mosaic 

 was created by delineating overlapping 

 zones on adjacent images based on the 

 relative positions of identifiable pieces of 

 ice. This mosaic allowed the analyst to 

 account for ice movement when counting 

 seals. In some cases, delineation of over- 

 lapping zones was difficult, particularly 



5 km 



