Get reliable science

Tracking collars utilizing GPS data loggers have become popular among cattle and rangeland researchers. These collars can be constructed at much lower cost than commercial collars, and a recent study has shown that metrics like distance traveled, elevation and slope are comparable. However, these lower-cost GPS collars do not come with software to identify and correct bad data. Past attempts to isolate bad data involved time consuming visual observation and removal using Geographic Information System (GIS) software. While this technique works well to find erroneous GPS point outside of pasture boundaries, points nested within the animal�s normal range are undetectable. Single metric, such as distance traveled, sorting on spreadsheet software often results in loss of good data surrounding bad points. In this study, authors characterized differences between commercially available GPS tracking collars (Lotek) and tracking collars made from lower-cost GPS data loggers (Knight), describe methods to isolate and remove bad GPS points from the data set, and looked at times of day when inaccurate data occurred more frequently.� Eleven Lotek and 6 Knight GPS tracking collars were placed on Gund Ranch cattle near Austin, Nevada for 140 d beginning 24 March 2016. Lotek collars had a fix rate of 86.9 %, missing 4.7% of scheduled observations, and inaccurate data points accounted for 0.54% of the observations while Knight collars were 73.1%, 26.4%, and 0.71%, respectively. An algorithm was developed to identify and remove bad data from Knight collar data sets using course change, rate of travel, and distance between points. For both Lotek and Knight collars, error frequency was highest in the morning before 0900 and at night around 2300.� This algorithm allows researchers to readily identify inaccurate positions and remove them from tracking data, which makes the Knight collars more useful for cattle tracking studies.