GPS Collar Sampling Frequency: Effects on Measures of Resource Use

Abstract

A challenge in animal behavior studies using Global Positioning System (GPS) collars is selecting a sampling frequency to accomplish desired goals. High data resolution (i.e., frequent sampling) is appealing, because it maximizes behavioral information garnered. Extended sampling might be needed, however, to describe long term behaviors or seasonal dynamics. Because tradeoffs exist between high data resolution and sampling duration, we evaluated the effects of variable GPS sampling intervals on proportions of pastures used by cattle and distance traveled per day. This was accomplished with GPS collars configured to record cattle positions every 5 min for 15 d in three 829-864-ha pastures. Data were iteratively reduced to simulate increasingly longer GPS recording intervals from once every 10 min up to once daily. Two techniques were used to measure the percentage of pastures accessed by cattle. The first counted only pixels containing GPS coordinates. The second counted pixels containing coordinates and/or traversed by lines between vertices. Expansion of GPS recording intervals decreased (P < 0.01) estimates of the proportion of pastures visited by cattle with rates of decline best fit by exponential decay functions for both line and point techniques (R2 = 0.93 and 0.97, respectively). Spatial errors accompanying less frequent sampling intervals, however, were extremely large with the line technique and misrepresented areas visited by cattle. Expansion of GPS sampling intervals decreased (P < 0.001) distance traveled per day by cattle about 10% with each iteration. If travel corridors or accurate assessments of resources accessed are of critical concern, then longer GPS integration intervals should be avoided because they propagate flawed spatial interpretations. Similarly, if accurate measures of travel distances are critical, we suggest using a relatively frequent GPS recording interval.