Northern Prairie Wildlife Research Center
Philip D. McLoughlin, Department of Biology, University of Saskatchewan, 112 Science Place, Saskatoon, SK S7N 5E2 Canada, Vivian Banci, 21557 Campbell Avenue, Maple Ridge, BC V2X 3V6 Canada, and H. Dean Cluff, Department of Resources, Wildlife, and Economic Development, Government of the Northwest Territories, P.O. Box 2668, Yellowknife, NT X1A 2P9 Canada
Previous estimates of error in satellite radio-collar locations have largely been based on stationary collars while little work has been done on estimating the error of collars attached to free-ranging study animals. Work on barren-ground grizzly bears (Ursus arctos) equipped with satellite radio-collars in the central Northwest Territories has shown that successful transmissions from stationary (dropped) collars to receiving satellites are greater than those attached to animals. Since the number of messages received by satellites contributes to accuracy and precision of location estimates, it was hypothesized that radio-collars deployed on bears will produce location estimates that are less accurate and precise than those from stationary collars (for any given Service Argos location class).
We tested our hypothesis by examining satellite-based locations of barren-ground grizzly bears in 1995 and 1996 that were simultaneously tracked using conventional (VHF) radio-tracking (radio-collars transmitted both as satellite and VHF beacons). On these and on other surveys, collared grizzly bears were directly observed from the air and the time recorded. A detailed site investigation was conducted of the area once the bear had traveled >5 km from the site. For this analysis we selected those locations associated with a stationary attractant such as kills, den entrances, and beds. A hand-held GPS receiver (Trimble Navigation with Acculoc function) was used to obtain location coordinates. We computed GPS-based locations by averaging approximately 250 fixes and therefore assume our non-differentially corrected fixes are accurate to about 20-30 m. We matched observed locations with those received from satellite telemetry. We compared locations that differed by no more than 4 hr unless we knew bears remained at a kill site for extended periods.
Our sample sizes remain small and we can only report collar means (± 1 S.E.) for locations with Service Argos classes 1 ( = 1471 m ± 233 m, n = 10) and 2 ( = 890 m ± 149 m, n = 9). These error estimates are considerably greater than previously reported errors for stationary collars; however, we recognize that our study animals may not have remained within only a few meters of kills, dens, or beds when the satellite location estimate occurred. Furthermore, our results are preliminary and we expect to increase our sample size over the next 2 years.