Research Document - 2008/064
An evaluation of Steller sea lion (Eumetopias jubatus) pup counts from 35mm oblique images
By P.F. Olesiuk, D.G. Calkins, K.W. Pitcher, W.L. Perryman, C. Stinchcomb, and M. Lynn
Abstract
The precision and accuracy of Steller sea lion (Eumetopias jubatus) pup counts made from oblique 35mm aerial slides was assessed by comparing them to ground drive-counts and counts from aerial vertical medium-format images. DFO flew surveys using oblique 35mm photography within 2 days of ADF&G’s ground drive-counts at Forrester Island, Alaska, in 1994, 1995, 1997 and 1998. In 1998 and 2002, aerial surveys were conducted at rookeries in B.C. and Forrester Island using both 35mm oblique and vertical medium-format photography, the latter of which was shown to provide pup counts statistically equivalent to ground drive-counts (Snyder et al. 2001). As expected, ground drive-counts provided the most precise pup counts (CV=0.047) and are widely regarded as the most accurate method against which other techniques are generally validated. In an earlier study, Snyder et al. (2001) showed that vertical medium-format images provided as good precision (CV=0.048) as ground counts on Alaskan rookeries. However, our initial counts of medium-format images for the B.C. survey in 1998 were less precise (CV=0.094) because one reader obtained counts that were consistently (8 of 10 sites) and significantly (0.0001<P<0.0110) greater than the other reader. The precision was improved (CV=0.056) by replacing the lower of the initial two reader’s counts with those of a third reader, whose counts agreed closely with the higher of the initial two readers. There were no significant differences among readers and fairly good precision (CV=0.063) for the medium-format images of B.C. rookeries in 2002, leading to an overall CV=0.060 for the two medium-format surveys. Counts made from oblique 35mm slides were reproducible among readers (CV=0.085), and were similar for surveys replicated on different dates (CV=0.102), but appeared to be slightly biased.
Comparison of pup counts from oblique 35mm slides to ground drive-counts at Forrester Island indicated that the 35mm counts tended to be significantly lower (P<0.0001), with about 80% of the pups seen on the ground evident in the 35mm slides. The degree of bias seemed relatively constant on a site-by-site basis (slope=0.797-0.813; 95% CI of 0.738-0.893) and between the 4 years (mean 79.7%; range 76-85%). Similarly, comparison of 35mm slide counts with medium-format counts at Forrester Island also indicated that only about 80% of pups seen in medium-format images were evident in the 35mm slides (slope=0.797-0.813; 95% CI of 0.693-0.891). The 35mm counts appeared to be less biased on B.C. rookeries, with about 96% of pups in the medium-format images being evident in the 35mm slides. We suspect that the difference in accuracy of 35mm pup counts between Forrrester Island and B.C. rookeries may be due to differences in the size and topography of rookeries. We therefore recommend that a correction factor of 1.25 (95% CI of 1.12-1.44) be applied to pup counts made from 35mm slides at Forrester Island, and a correction factor of 1.05 (95% CI of 1.018-1.075) be applied to pup counts made from 35mm slides at B.C. rookeries, to account for pups that are obscured in photographs taken at oblique angles.
In 2006, we began a transition from 35mm slides to digital photography. To insure consistency of the survey time-series, all rookeries in B.C. where photographed using both a film and digital SLR camera. There was close agreement between the digital and film counts for both pups (0.9942<r2<0.9954) and non-pups (0.9994<r2=0.9994), and in both cases the intercepts passed through the origin (0.423<P<0.742) and slopes were not significantly different from unity (0.503<P<0.849) indicating the counts were statistically indistinguishable from one another.
Accessibility Notice
This document is available in PDF format. If the document is not accessible to you, please contact the Secretariat to obtain another appropriate format, such as regular print, large print, Braille or audio version.
- Date modified: