Submitted to: Sensors
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: March 12, 2013
Publication Date: March 13, 2013
Citation: Augustine, D.J., Derner, J.D. 2013. Assessing herbivore foraging behavior with GPS collars in a semiarid grassland. Sensors. 13:3711-3723. Interpretive Summary: We conducted a study to examine how accurately one can predict the behavior of free-ranging cattle from information collected by collars that (1) record the animal’s location (using global positioning systems [GPS]) and (2) contain an activity sensor that records information about the animal’s head movements. We collected GPS collar data and direct observations of cattle behavior for four years (2008-2011) in semi-arid rangeland of eastern Colorado. First, we examined how well we could distinguish between grazing versus non-grazing behaviors. Second, we examined how well we could distinguish among four categories of behavior: grazing, resting, travelling, and a ‘mixed’ activity category. In the first analysis, our best model was able to distinguish between grazing versus non-grazing behavior 87.1% of the time (corresponding to a misclassification rate of 12.9%). For the multi-category analysis, we the misclassification rate increased to 16.4%. The distance that the animal travelled in a 5-min interval and the proportion of the interval with the sensor in the head down position were the two most important variables predicting grazing activity. Our results differ from previous assessments for rangleands of Israel and for pastures in the United States (Kentucky) in terms of the value of different activity sensor measurements for identifying grazing activity, indicating that the use of GPS collars to classify cattle behavior will require calibrations specific to the environment and vegetation being studied.
Technical Abstract: Advances in global positioning system (GPS) technology have dramatically enhanced the ability to track and study distributions of free-ranging livestock. However, understanding factors controlling livestock foraging distribution requires the ability to assess when and where they are foraging. We collected synchronized GPS collar data and direct observations of cattle behavior for four years (2008-2011) in semi-arid rangeland of eastern Colorado, and developed classification tree models discriminating between grazing and non-grazing activity. We evaluated (1) which activity sensor measurements from the GPS collars were most valuable in predicting cattle foraging behavior, (2) the accuracy of binary (grazing, non-grazing) activity models versus models with multiple activity categories (grazing, resting, traveling, mixed), and (3) the accuracy of models that are robust across years versus models specific to a given year. A binary classification tree correctly removed 86.5% of the non-grazing locations, while correctly retaining 87.8% of the locations where the animal was grazing, for an overall misclassification rate of 12.9%. A classification tree with four activity categories yielded a misclassification rate of 16.0%. The distance travelled in a 5-min interval and the proportion of the interval with the sensor in the head down position were the two most important variables predicting grazing activity. Fitting annual models of cattle foraging activity did not improve model accuracy compared to a single model based on all years combined, indicating that increased sample size was more valuable than accounting for interannual variation in foraging behavior. Our models differ from previous assessments in semi-arid rangeland of Israel [1,2] and mesic pastures in the United States  in terms of the value of different activity sensor measurements for identifying grazing activity, indicating that the use of GPS collars to classify cattle grazing behavior will require calibrations specific to the environment and vegetation being studied.