Ronald Aaron Moen

Center for Water and the Environment    .    Natural Resources Research Institute

GPS Collar Research

Moen, R.A., J. Pastor, and Y. Cohen. 2001. Effect of animal movement on GPS telemetry locations. Alces 37:207-271. Click here for 667K pdf

Interpretation of habitat use from GPS collar locations could be biased if the activity of animals wearing GPS collars affects the probability of obtaining a successful location. We tested for this bias with GPS location attempts made by collars deployed on free-ranging moose ( Alces alces ) in northern Minnesota, USA. We classified moose as being either inactive or active during each GPS location attempt based on activity counts recorded by the collar. Only 69% of GPS location attempts were successful while moose were active, compared to 88% when moose were inactive. Moose activity reduced success of location attempts in both summer and winter. We also estimated the precision of GPS locations while collars were deployed on free-ranging moose. When moose were inactive 50% of 3-dimensional locations were within 5 m of the estimated location, and 95% were within 17 m of the estimated location. When moose were inactive, 50% of 2-dimensional locations were within 7 m of the estimated location, and 95% were within 26 m of the estimated location. Despite the bias induced by animal activity, GPS telemetry is the most precise method currently available to obtain locations of free-ranging large mammals such as moose. Sampling biases in GPS units resulting from animal activity should be accounted for when interpreting habitat use by free-ranging animals.

Moen R.A., J. Pastor, Y. Cohen, and C.C. Schwartz. 1996. Effects of moose movement and habitat use on GPS collar performance. Journal of Wildlife Management 60:659-668.

We tested a radiotelemetry collar that uses a Global Positioning System (GPS) unit to calculate animal locations. We placed the collar in a range of cover types and compared locations reported by the collar to differentially corrected GPS locations. We placed the collar on a free ranging moose (Alces alces) and determined how selection of cover types, collar movement, and collar orientation affected GPS locations. On or off the moose, the GPS unit collected a location in >90% of location attempts in areas with no or thin canopy cover, including mature deciduous canopies in winter. Under a mature conifer canopy or a mature deciduous canopy in summer, 60 to 70% of location attempts were successful. Locations from the GPS unit in the collar were close to the expected precision of non differentially corrected GPS (within 40 m 50% of the time and 100 m 95% of the time). Locations did not have a directional bias. Movement of the moose while a location was being attempted did not affect GPS locations. The moose occasionally laid down so the collar was horizontal. Although this decreased the success of location attempts, <1% of location attempts were so affected. GPS radiotelemetry has great promise for expanding our knowledge about hourly, daily, and annual patterns in animal movements and habitat selection.

Moen, R.A., J. Pastor, and Y. Cohen. 1997. Accuracy of GPS telemetry collar locations with differential correction. Journal of Wildlife Management 61:530-539.

Global Positioning System (GPS) units in telemetry collars provide an unbiased and precise estimate of animal locations, Under ideal conditions at least 50% of locations are expected to be within 40 m in uncorrected mode GPS, and within 5 m in differential mode GPS. When the collar was placed under open sky, most locations were 3 dimensional locations that could be differentially corrected. Under hardwood canopies with leaves on, the frequency of 3 dimensional locations decreased: the frequencies of failed location attempts and 2 dimensional locations increased, and the precision of GPS locations decreased. We compared the precision of each GPS mode by calculating uncorrected mode and differential mode locations from the same pseudo range and ephemeris data. We varied the number of satellites used in the location solution to simulate the effect of decreased satellite acquisition due to canopy cover On precision of locations. Precision of locations increased if signals from >4 satellites were used to calculate tile location in uncorrected mode and in differential mode. We found that 2 dimensional locations Here almost as precise as 3 dimensional positions if the altitude of the GPS unit was known. If the altitude used to calculate a 2 dimensional location was within 50 m of the actual collar altitude. the precision of 2 dimensional differential mode locations was better than 3 dimensional uncorrected mode locations, If the error in altitude was 100 or 150 m, then 50% of 2 dimensional differential mode locations were within 70 m and 95% were within 185 m of the true location. We used GPS locations from collars placed in different cover types and on free ranging moose (Alces alces) to determine the effect of season, time of day, rainfall, and cover type on CPS performance. On free ranging moose the collar GPS unit found greater than or equal to 4 satellites on 52% of location attempts. >50% of locations were 3 dimensional, and >24% of locations were e dimensional. Precise tracking of individual animals in all weather throughout the ear is possible with GPS telemetry.

Moen, R.A., J. Pastor, and Y. Cohen. 1996. Interpreting behavior from activity counters in GPS collars on moose. Alces 32:101-108. Click here for 118K pdf

Activity patterns of free-ranging moose can be estimated from activity counts on radiotelemetry collars. We observed a collared moose to calibrate activity counts on a collar to activities of moose in natural habitats. If activity counts were low, there was a high probability that the moose was inactive, but we recorded high activity counts when the moose was active and on 25% of time intervals when the moose was inactive. We believe the high activity counts were due to collar placement and due to bugs in the software controlling the activity counter. We also analyzed activity counts from collars on free-ranging moose which were not observed. There were cyclic periods of activity and inactivity throughout the day when activity counts were taken every 10 minutes on 6 free-ranging moose. We averaged activity counts taken on 10-minute intervals to simulate activity counts reported by the collar when intervals between GPS locations were 1 to 4 hours. The correspondence between averaged activity counts and percent of time spent active each day was poor if many activity counts were averaged. It is not possible to interpret daily activity patterns of moose from activity counts which are averaged over periods longer than 1 hour.

University of Minnesota-Duluth    .    5013 Miller Trunk Highway    .    Duluth, MN 55811

Voice: (218)720-4372    .    Fax: (218)720-9412    .    Email: rmoen@nrri.umn.edu